IEEE News

Learn about carrier aggregation, microcell overlapping, and massive MIMO implementation. Delve into the world of surface acoustic wave (SAW) and bulk acoustic wave (SAW) filters and understand their strengths, limitations, and applications in the evolving 5G/6G landscape. Key highlights: Uncover the design challenges of new technologies in the mobile ecosystem Explore the field of SAW and SAW filters and discover their roles and performance nuances Take a look at the impact of temperature on filter technologies and how it shapes their applications Learn how simulation technology can bridge the gap between design concepts and real-world implementation Stay ahead in 5G/6G innovation and learn how filters shape seamless communication. Register now for this free webinar!

In the latest round of machine learning benchmark results from MLCommons, computers built around Nvidia’s new Blackwell GPU architecture outperformed all others. But AMD’s latest spin on its Instinct GPUs, the MI325, proved a match for the Nvidia H200, the product it was meant to counter. The comparable results were mostly on tests of one of the smaller-scale large language models, Llama2 70B (for 70 billion parameters). However, in an effort to keep up with a rapidly changing AI landscape, MLPerf added three new benchmarks to better reflect where machine learning is headed. MLPerf runs benchmarking for machine learning systems in an effort to provide an apples-to-apples comparison between computer systems. Submitters use their own software and hardware, but the underlying neural networks must be the same. There are a total of 11 benchmarks for servers now, with three added this year. It has been “hard to keep up with the rapid development of the field,” says Miro Hodak, the cochair of MLPerf Inference. ChatGPT appeared only in late 2022, OpenAI unveiled its first large language model (LLM) that can reason through tasks last September, and LLMs have grown exponentially—GPT3 had 175 billion parameters, while GPT4 is thought to have nearly 2 trillion. As a result of the breakneck innovation, “we’ve increased the pace of getting new benchmarks into the field,” says Hodak. The new benchmarks include two LLMs. The popular and relatively compact Llama2 70B is already an established MLPerf benchmark, but the consortium wanted something that mimicked the responsiveness people are expecting of chatbots today. So the new benchmark “Llama2-70B Interactive” tightens the requirements. Computers must produce at least 25 tokens per second under any circumstance and cannot take more than 450 milliseconds to begin an answer. Seeing the rise of “agentic AI”—networks that can reason through complex tasks—MLPerf sought to test an LLM that would have some of the characteristics needed for that. They chose Llama3.1 405B for the job. That LLM has what’s called a wide context window. That’s a measure of how much information—documents, samples of code, et cetera—it can take in at once. For Llama3.1 405B, that’s 128,000 tokens, more than 30 times as much as Llama2 70B. The final new benchmark, called RGAT, is what’s called a graph attention network. It acts to classify information in a network. For example, the dataset used to test RGAT consists of scientific papers, which all have relationships between authors, institutions, and fields of study, making up 2 terabytes of data. RGAT must classify the papers into just under 3,000 topics. Blackwell, Instinct Results https://public.flourish.studio/visualisation/22428...” width=”100%”> Nvidia continued its domination of MLPerf benchmarks through its own submissions and those of some 15 partners, such as Dell, Google, and Supermicro. Both its first- and second-generation Hopper architecture GPUs—the H100 and the memory-enhanced H200—made strong showings. “We were able to get another 60 percent performance over the last year” from Hopper, which went into production in 2022, says Dave Salvator, director of accelerated computing products at Nvidia. “It still has some headroom in terms of performance.” But it was Nvidia’s Blackwell architecture GPU, the B200, that really dominated. “The only thing faster than Hopper is Blackwell,” says Salvator. The B200 packs in 36 percent more high-bandwidth memory than the H200, but, even more important, it can perform key machine learning math using numbers with a precision as low as 4 bits instead of the 8 bits Hopper pioneered. Lower-precision compute units are smaller, so more fit on the GPU, which leads to faster AI computing. In the Llama3.1 405B benchmark, an eight-B200 system from Supermicro delivered nearly four times the tokens per second of an eight-H200 system by Cisco. And the same Supermicro system was three times as fast as the quickest H200 computer at the interactive version of Llama2 70B. Nvidia used its combination of Blackwell GPUs and Grace CPU, called GB200, to demonstrate how well its NVL72 data links can integrate multiple servers in a rack, so they perform as if they were one giant GPU. In an unverified result the company shared with reporters, a full rack of GB200-based computers delivers 869,200 tokens per second on Llama2 70B. The fastest system reported in this round of MLPerf was an Nvidia B200 server that delivered 98,443 tokens per second. AMD is positioning its latest Instinct GPU, the MI325X, as providing performance competitive with Nvidia’s H200. MI325X has the same architecture as its predecessor, MI300, but it adds even more high-bandwidth memory and memory bandwidth—256 gigabytes and 6 terabytes per second (a 33 percent and 13 percent boost, respectively). Adding more memory is a play to handle larger and larger LLMs. “Larger models are able to take advantage of these GPUs because the model can fit in a single GPU or a single server,” says Mahesh Balasubramanian, director of data-center GPU marketing at AMD. “So you don’t have to have that communication overhead of going from one GPU to another GPU or one server to another server. When you take out those communications, your latency improves quite a bit.” AMD was able to take advantage of the extra memory through software optimization to boost the inference speed of DeepSeek-R1 eightfold. On the Llama2 70B test, an eight-GPU MI325X computers came within 3 to 7 percent the speed of a similarly tricked-out H200-based system. And on image generation the MI325X system was within 10 percent of the Nvidia H200 computer. AMD’s other noteworthy mark this round was from its partner, Mangoboost, which showed nearly fourfold performance on the Llama2 70B test by doing the computation across four computers. Intel has historically put forth CPU-only systems in the inference competition to show that for some workloads you don’t really need a GPU. This time around saw the first data from Intel’s Xeon 6 chips, which were formerly known as Granite Rapids and are made using Intel’s 3-nanometer process. At 40,285 samples per second, the best image-recognition results for a dual-Xeon 6 computer was about one-third the performance of a Cisco computer with two Nvidia H100s. Compared with Xeon 5 results from October 2024, the new CPU provides about an 80 percent boost on that benchmark and an even bigger boost on object detection and medical imaging. Since it first started submitting Xeon results in 2021 (the Xeon 3), the company has achieved an elevenfold boost in performance on Resnet. For now, it seems Intel has quit the field in the AI accelerator-chip battle. Its alternative to the Nvidia H100, Gaudi 3, did not make an appearance in the new MLPerf results, nor in version 4.1, released last October. Gaudi 3 got a later-than-planned release because its software was not ready. In the opening remarks at Intel Vision 2025, the company’s invite-only customer conference, newly minted CEO Lip-Bu Tan seemed to apologize for Intel’s AI efforts. “I’m not happy with our current position,” he told attendees. “You’re not happy either. I hear you loud and clear. We are working toward a competitive system. It won’t happen overnight, but we will get there for you.” Google’s TPU v6e chip also made a showing, though the results were restricted to the image-generation task. At 5.48 queries per second, the 4-TPU system saw a 2.5-times boost over a similar computer using its predecessor TPU v5e in the October 2024 results. Even so, 5.48 queries per second was roughly in line with a similarly sized Lenovo computer using Nvidia H100s. This post was corrected on 2 April 2025 to give the right value for high-bandwidth memory in the MI325X.

In particle physics, the smallest problems often require the biggest solutions. Along the border of France and Switzerland, around a hundred meters underneath the countryside, protons speed through a 27-kilometer ring—about seven times the length of the Indy 500 circuit—until they crash into protons going in the opposite direction. These particle pileups produce a petabyte of data every second, the most interesting of which is poured into data centers, accessible to thousands of physicists worldwide. The Large Hadron Collider (LHC), arguably the largest experiment ever engineered, is needed to probe the universe’s smallest constituents. In 2012, two teams at the LHC discovered the elusive Higgs boson, the particle whose existence confirmed 50-year-old theories about the origins of mass. It was a scientific triumph that led to a Nobel Prize and worldwide plaudits. Since then, experiments at the LHC have focused on better understanding how the newfound Higgs fits into the Standard Model, particle physicists’ best theoretical description of matter and forces—minus gravity. “The Standard Model is beautiful,” says Victoria Martin, an experimental physicist at the University of Edinburgh. “Because it’s so precise, all the little niggles stand out.” The Large Hadron Collider lives in a 27-kilometer tunnel ring, about 100 meters underneath France and Switzerland. It was used to discover the Higgs boson, but further research may require something larger still. Maximilien Brice/CERN The minor quibbles physicists have about the Standard Model could be explained by new particles: Dark matter, the invisible material whose gravity shapes the universe, is thought to be made of heretofore undiscovered particles. But such new particles may be out of reach for the LHC, even after it undergoes upgrades that are set to be completed later this decade. To address these lingering questions, particle physicists have been planning its successors. These next-generation colliders will improve on the LHC by smashing protons at higher energies or by making more precise collisions with muons, antimuons, electrons, and positrons. In doing so, they’ll allow researchers to peek into a whole new realm of physics. Martin herself is particularly interested in learning more about the Higgs, and learning exactly how the particle responsible for mass behaves. One possible find: Properties of the Higgs suggest that the universe might not be stable in the long, long term. [Editor’s note: About 10790 years. Other problems may be more pressing.] “We don’t really know exactly what we’re going to find,” Martin says. “But that’s okay, because it’s science, it’s research.” There are four main proposals for new colliders, and each one comes with its own slew of engineering challenges. To build them, engineers would need to navigate tricky regional geology, design accelerating cavities, handle the excess heat within the cavities, and develop powerful new magnets to whip the particles through these cavities. But perhaps more daunting are the geopolitical obstacles: coordinating multinational funding commitments and slogging through bureaucratic muck. Collider projects take years to plan and billions of dollars to finance. The fastest that any of the four machines would come on line is the late 2030s. But now is when physicists and engineers are making key scientific and engineering decisions about what’s coming next. Supercolliders at a glance Large Hadron Collider Size (circumference): 27 kilometers Collision energy: 13,600 giga-electron volts Colliding particles: protons and ions Luminosity: 2 × 1034 collisions per square centimeter per second (5 × 1034 for high-luminosity upgrade) Location: Switzerland–France border Start date: 2008– International Linear Collider Size (length): 31 km Collision energy: 500 GeV Colliding particles: electrons and positrons Luminosity (at peak energy): 3 × 1034 collisions per cm2 per second Location: Iwate, Japan Earliest start date: 2038 Muon collider Size (circumference): 4.5 km (or 10 km) Collision energy: 3,000 GeV (or 10,000 GeV) Colliding particles: muons and antimuons Luminosity: 2 × 1035 collisions per cm2 per second Location: possibly Fermilab Earliest start date: 2045 (or in the mid-2050s) Future Circular Collider-ee | FCC-hh Size (circumference): 91 km Collision energy: 240 GeV | 85,000 GeV Colliding particles: electrons and positrons | protons Luminosity: 8.5 × 1034 | 30 × 1034 collisions per cm2 per second Location: Switzerland–France border Earliest start date: 2046 | 2070 Circular Electron Positron Collider | Super proton–proton Collider (SPPC) Size (circumference): 100 km Collision energy: 240 GeV | 100,000 GeV Colliding particles: electrons and positrons | protons Luminosity: 8.3 × 1034 | 13 × 1034 collisions per cm2 per second Location: China Earliest start date: 2035 | 2060s Possible supercolliders of the future The LHC collides protons and other hadrons. Hadrons are like beanbags, full of quarks and gluons, that spray around everywhere upon collision. Next-generation colliders have two ways to improve on the LHC: They can go to higher energies or higher precision. Higher energies provide more data by producing more particles—potentially new, heavy ones. Higher-precision collisions give physicists cleaner data with a better signal-to-noise ratio because the particle crash produces less debris. Either approach could reveal new physics beyond the Standard Model. Three of the new colliders would improve on the LHC’s precision by colliding electrons and their antimatter counterparts, positrons, instead of hadrons. These particles are more like individual marbles—much lighter, and not made up of any smaller constituents. Compared with the collisions between messy, beanbag-like hadrons, a collision between electrons and positrons is much cleaner. After taking data for years, some of those colliders could be converted to smash protons as well, though at energies about eight times as high as those of the LHC. These new colliders range from technically mature to speculative. One such speculative option is to smash muons, electrons’ heavier cousins, which have never been collided before. In 2023, an influential panel of particle physicists recommended that the US pursue development of such a machine, in a so-called ‘muon shot’. If it is built, a muon collider would likely be based at Fermilab, the center of particle physics in the United States. A muon collider “can bring us outside of the world that we know,” says Daniele Calzolari, a physicist working on muon collider design at CERN, the European Organization for Nuclear Research. “We don’t know exactly how everything will look like, but we believe we can make it work.” While muon colliders have remained conceptual for more than 50 years, their potential has long excited and intrigued physicists. Muons are heavy compared with electrons, almost as heavy as protons, but they lack the mess of quarks and gluons, so collisions between muons could be both high energy and high precision. Superconducting radio-frequency cavities are used in particle colliders to apply electric fields to charged particles, speeding them up toward each other until they smash together. Newer methods of making these cavities are seamless, providing more-precise steering and, presumably, better collisions. Reidar Hahn/Fermi The trouble is that muons decay rapidly—in a mere 2.2 microseconds while at rest—so they have to be cooled, accelerated, and collided before they expire. Preliminary studies suggest a muon collider is possible, but key technologies, like powerful high-field solenoid magnets used for cooling, still need to be developed. In March 2025, Calzolari and his colleagues submitted an internal proposal for a preliminary demonstration of the cooling technology, which they hope will happen before the end of the decade. The accelerator that could theoretically come on line the soonest, would be the International Linear Collider (ILC) in Iwate, Japan. The ILC would send electrons and positrons down straight tunnels where the particles would collide to produce Higgs bosons that are easier to detect than at the LHC. The collider’s design is technically mature, so if the Japanese government officially approved the project, construction could begin almost immediately. But after multiple delays by the government, the ILC remains in a sort of planning purgatory, looking more and more unlikely. The Standard Model of particle physics is the current best theory of all the understood matter and forces in our universe (except gravity). The model works extremely well, but scientists also know that it is incomplete. The next generation of supercolliders might give a glimpse at what’s beyond the Standard Model. So, the two colliders, which are both technically mature, that have perhaps the clearest path to construction are China’s Circular Electron Positron Collider (CEPC) and CERN’s Future Circular Collider (FCC-ee). CERN’s FCC-ee would be a 91-km ring, designed to initially collide electrons and positrons to study the parameters of particles like the Higgs in fine detail (the “ee” indicates collisions between electrons and positrons). Compared with the LHC’s collisions of protons or heavy ions, those between electrons and positrons “are much cleaner, so you can have a more precise measurement,” says Michael Benedikt, the head of the FCC-ee effort. After about a decade of operation—enough time to gather data and develop the needed magnets—it would be upgraded to collide protons and search for new physics at much higher energies (and then become known as the FCC-hh, for hadrons). The FCC-ee’s feasibility report just concluded, and CERN’s member states are now left deciding whether to pursue the project. Similarly, China’s CEPC would also be a 100-km ring designed to collide electrons and positrons for the first 18 years or so. And much like the FCC, a proton or other hadron upgrade is in the works after that. Later this year, Chinese researchers plan to submit the CEPC for official approval by the Chinese government as part of the next five-year-plan. As the two colliders (and their proton upgrades) are considered for construction in the next few years, policymakers will be thinking about more than just their potential for discovery. CEPC and FCC-ee are, in this sense, less abstract physics experiments and more engineering projects with concrete design challenges. Laying the groundwork When particles zip around the curve of a collider, they lose energy—much like a car braking on a racetrack. The effect is particularly pronounced for lightweight particles like electrons and positrons. To reduce this energy loss from sharp turns, CEPC and FCC-ee are both planned to have enormous tunnels, which, if built, would be among the longest in the world. The construction cost of such an enormous tunnel would be several billion U.S.dollars, roughly one-third of the total collider price. Finding a place to bury a 90-km ring is not easy, especially in Switzerland. The proposed path of the FCC-ee has an average depth of 200 meters, with a dip to 500 meters under Lake Geneva, fit snugly between the Jura Mountains to the northwest and the Prealps to the east. The land there was once covered by a sea, which left behind sedimentary rock—a mixture of sandstone and shale known as molasse. “We’ve done so much tunneling at CERN before. We were quite confident about the molasse rock,” says Liam Bromiley, a civil engineer at CERN. But the FCC-ee’s path also takes it through deposits of limestone, which is permeable and can hold karsts, or cavities, full of water. “If you hit one of those, you could end up flooding the tunnel,” Bromiley says. During the next two years, if the project is green-lit, engineers will drill boreholes into the limestone to determine whether there are karsts that can be avoided. FCC-ee would be a 91-km ring spanning underneath Switzerland and France, near the current Large Hadron Collider. One of the proposed locations for the CEPC is near the northern port city of Qinhuangdao, where the 100 km circumference collider would be buried underground.Chris Philpot CEPC, in contrast, has a much looser spatial constraint, and can choose from nearly anywhere in China. Three main sites are being considered: Qinhuangdao (a northern port city), Changsha (a metropolis in central China), and Huzhou (a coastal city near Shanghai). According to Jie Gao, a particle physicist at the Institute of High Energy Physics, in Beijing, the ideal location will have hard rock, like granite, and low seismic activity. Additionally, Gao says, they want a site with good infrastructure to create a “science city” ideal for an international community of physicists. The colliders’ carbon footprints are also on the minds of physicists. One potential energy-saving measure: redirecting excess heat from operations. “In the past we used to throw it into the atmosphere,” Benedikt says. In recent years, heated water from one of the LHC’s cooling stations has kept part of the commune of Ferney-Voltaire warm during the winters, and Benedikt says the FCC-ee would expand these environmental efforts. Getting up to speed If the civil-engineering challenges are met, physicists will rely on a spate of technologies to accelerate, focus, and collide electrons and positrons at CEPC and FCC-ee more precisely and efficiently than they could at the LHC. When both types of particles are first produced from their sources, they start off at a comparatively low energy, around 4 giga-electron volts. To get them up to speed, electrons and positrons are sent through superconducting radio-frequency (SRF) cavities—gleaming metal bubbles strung together like beads of a necklace, which apply an electric field that pushes the charged particles forward. Both China’s Circular Electron Positron Collider (CEPC) [bottom] and CERN’s Future Circular Collider (FCC-ee) [top] have preliminary designs of the insides of their tunnels, including the collider itself, associated vacuum and control equipment, and detectors.Chris Philpot In the past, SRF cavities were welded together, which inherently left imperfections that led to beam instabilities. “You can never obtain a perfect surface along this weld,” Benedikt says. FCC-ee researchers have explored several techniques to create cavities without seams, including hydroforming, which is widely used for the components of high-end sports cars. A metal tube is placed in a pressurized cell and compressed against a die by liquid. The resulting cavity has no seams and is smooth as blown glass. To improve efficiency, engineers focus on the machines that power the SRF cavities, machines called klystrons. Klystrons have historically had efficiencies that peak around 65 percent, but design advances, such as the machines’ ability to bunch electrons together, are on track to reach efficiencies of 80 percent. “The efficiency of the klystron is becoming very important,” Gao says. Over 10 years of operation, these savings could amount to 1 terawatt hour—about enough electricity to power all of China for an hour. Another efficiency boost comes from focusing on the tunnel design. As electrons and positrons follow the curve of the ring, they will lose a considerable amount of energy, so SRF cavities will be placed around the ring to boost particle energies. The lost energy will be emitted as potent synchrotron radiation—about 10,000 times as much radiation as is emitted by protons circling the LHC today. “You do not want to send the synchrotron radiation into the detectors,” Benedikt says. To avoid this fate, neither FCC-ee nor CEPC will be perfectly circular. Shaped a bit like a racetrack, both colliders will have about 1.5-km-long straight sections before an interaction point. Other options are also on the table—in the past, researchers have even used repurposed steel from scrapped World War II battleships to shield particle detectors from radiation. Both CEPC and FCC-ee will be massive data-generating machines. Unlike the LHC, which is regularly stopped to insert new particles, the next-generation colliders will be fed with a continuous stream of particles, allowing it to stay in “collision mode” and take more data. At a collider, data is a function of ‘luminosity’— the ratio of detected events per square centimeter, per second. The more particle collisions, the “brighter” the collider. Firing particles at each other is a little like trying to get two bullets to collide—they often miss each other, which limits the luminosity. But physicists have a variety of strategies to squeeze more electrons and positrons into smaller areas to achieve more of these unlikely collisions. Compared to the Large Electron-Positron (LEP) collider of the 1990s, the new machines will produce 100,000 times as many Z bosons—particles responsible for radioactive decay. More Z bosons means more data. “The FCC-ee can produce all the data that were accumulated in operation over 10 years of LEP within minutes,” Benedikt says. Back to protons While both the FCC-ee and CEPC would start with electrons and positrons, they are designed to eventually collide protons. These upgrades are called FCC-hh and Super proton-proton Collider (SPPC). Using protons, FCC-hh and SPPC would reach a collision energy of 100,000 GeV, roughly an order of magnitude higher than the LHC’s 13,600 GeV. Though the collisions would be messy, their high energy would allow physicists to “explore fully new territory,” Benedikt says. While there’s no guarantee, physicists hope that territory teems with discoveries-in-waiting, such as dark-matter particles, or strange new collisions where the Higgs recursively interacts with itself many times. One pro of protons is that they are over 1,800 times as heavy as electrons, so they emit far less radiation as they follow the curve of the collider ring. But this extra heft comes with a substantial cost: Bending protons’ paths requires even stronger superconducting magnets. Magnet development has been the downfall of colliders before. In the early 1980s, a planned collider named Isabelle was scrapped because magnet technology was not far enough along. The LHC’s magnets are made from a strong alloy of niobium-titanium, wound together into a coil that produces magnetic fields when subjected to a current. These coils can produce field strengths over 8 teslas. The strength of the magnet pushes its two halves apart with a force of nearly 600 tons per meter. “If you have an abrupt movement of the turns in the coil by as little as 10 micrometers,” the entire magnet can fail, says Bernhard Auchmann, an expert on magnets at CERN. It is unlikely that any Collider—whether based in China, at CERN, the United States, or Japan—will be able to go it alone. Future magnets for FCC-hh and SPPC will need to have at least twice the magnetic field strength, about 16 to 20 T, pushing the limits of materials and physics. Auchmann points to three possible paths forward. The most straightforward option might be “niobium three tin” (Nb3Sn). Substituting tin for titanium allows the metal to host magnetic fields up to 16 T but makes it quite brittle, so you can’t “clamp the hell out of it,” Auchmann says. One possible solution involves placing (Nb3Sn) into a protective steel endoskeleton that prevents it from crushing itself. Then there are high-temperature superconductors. Some magnets made with rare earth metals can exceed 20 T, but they too are fragile and require similar steel supports. Currently, these materials are expensive, but demand from fusion startups, which also require these types of magnets, may push the price down, Auchmann says. Finally, there is a class of iron-based high-temperature superconductors that is being championed by physicists in China, thanks to the low price of iron and manufacturing-process improvements. “It’s cheap,” Gao says. “This technology is very promising.” Over the next decade or so, physicists will work on each of these materials, and hope to settle on one direction for next-generation magnets. Time and money While FCC-ee and CEPC (as well as their proton upgrades) share many of the same technical specifications, they differ dramatically in two critical factors: timelines and politics. Construction for CEPC could begin in two years; the FCC-ee would need to wait about another decade. The difference comes down largely to the fact that CERN has a planned upgrade to the LHC—enabling it to collect 10 times as much data—which will consume resources until nearly 2040. China, by contrast, is investing heavily in basic research and has the funds immediately at hand. The abstruse physics that happens at colliders is never as far from political realities on Earth as it seems. Japan’s ILC is in limbo because of budget issues. The muon collider is subject to the whims of the highly divided 119th U.S. Congress. Last year, a representative for Germany criticized the FCC-ee for being unaffordable, and CERN continues to struggle with the politics of including Russian scientists. Tensions between China and the United States are similarly on the rise following the Trump administration’s tariffs. How physicists plan to tackle these practical problems remains to be seen. But it is unlikely that any collider—whether based in China, at CERN, the United States, or Japan—will be able to go it alone. In addition to the tens of billions of dollars for construction and operation of the new facility, the physics expertise needed to run it and perform complex experiments at scale must be global. “By definition, it’s an international project,” Gao says. “The door is wide open.”

Most haptic interfaces today are limited to simple vibrations. While visual displays and audio systems have continued to progress, those using our sense of touch have largely stagnated. Now, researchers have developed a haptics system that creates more complex tactile feedback. Beyond just buzzing, the device simulates sensations like pinching, stretching, and tapping for a more realistic experience. “The sensation of touch is the most personal connection that you can have with another individual,” says John Rogers, a professor at Northwestern University in Evanston, Ill., who led the project. “It’s really important, but it’s much more difficult than audio or video.” Co-led by Rogers and Yonggang Huang, also a professor at Northwestern, the work is largely geared toward medical applications. But the technology could be used in a wide range of uses, including virtual or augmented reality and the ability to feel the texture of clothing fabric or other items while shopping online. The research was published in the journal Science on 27 March. A Nuanced Sense of Touch Today’s haptic interfaces mostly rely on vibrating actuators, which are fairly simple to construct. “It’s a great place to start,” says Rogers. But going beyond vibration could help add the vibrancy of real-world interactions to the technology, he adds. These types of interactions require more-sophisticated mechanical forces, which include a combination of both normal forces directed perpendicular to the skin’s surface and shear forces directed parallel to the skin. Whether through vibration or applying pressure, forces directed vertically into the skin have been the main focus of haptic designs, according to Rogers. But these don’t fully engage the many receptors embedded in our skin. The researchers aimed to build an actuator that offers full freedom of motion, which they achieved with “very old physics,” Rogers says—namely, electromagnetism. The basic design of the device consists of three nested copper coils and a small magnet. Running current through the coils generates a magnetic field that then moves the magnet, which delivers force to the skin. “What we’ve put together is an engineering embodiment [of the physics] that provides a very compact force delivery system and offers full programmability in direction, amplitude, and temporal characteristics,” says Rogers. For a more elaborate setup, the researchers also developed a version that uses a collection of four magnets with different orientations of north and south poles. This creates even more complex sensations of pinching, stretching, and twisting. Haptics at Your Fingertips—or Anywhere Because fingertips are highly sensitive, only small forces are needed for this application. John A. Rogers/Northwestern University Although much of the previous work in haptics has focused on fingertips and the hands, these devices could be placed elsewhere on the body, including the back, chest, or arms. However, these applications may have different requirements. Compared with places like the back, the fingertips are highly sensitive—both in terms of the force needed and the spatial density of receptors. “The fingertips are probably the most challenging in terms of density, but they’re easiest in terms of the forces that you need to deliver,” says Rogers. In other use cases, delivering enough power may be a challenge, he acknowledges. The force possible may also be limited by the size of the coils, says Gregory Gerling, a systems engineering professor at the University of Virginia and former chair of the IEEE Technical Committee on Haptics. The coil size dictates how much force you can generate, and at a certain point, the device won’t be wearable. However, he believes it is sufficient for VR applications. Gerling, an IEEE senior member, finds the use of magnetism in multiple directions interesting. Compared with other approaches that are based on hydraulics or air pressure, this system doesn’t require pumping fluids or gases. “You can be kind of untethered,” Gerling says. “Overall, it’s a very interesting, novel device, and maybe it takes the field in a slightly new direction.” Applications in VR, Neuropathy, and More The clearest application of the device is probably in virtual or augmented reality, says Rogers. These environments now have highly sophisticated audio and video inputs, “but the tactile component of that experience is still a work in progress,” he says. Their lab, however, is primarily focused on medical applications, including sensory substitution for patients who have lost sensation in a part of the body. A complex haptics interface could reproduce the sensation in another part of the body. For example, nerve damage in people with diabetic neuropathy makes it difficult for them to walk without looking at their feet. The lab is experimenting with placing an array of pressure sensors into the base of these patients’ shoes, then reproducing the pattern of pressure using a haptic array mounted on their upper thighs, where they still have sensation. The researchers are working with a rehabilitation facility in Chicago to test the approach, mainly with this population. Continuing to develop these medical applications will be a focus moving forward, says Rogers. In terms of engineering, he would like to further miniaturize the actuators to make dense arrays possible in regions of the body like the fingertips. Feeling the Music Additionally, the researchers explored the possibility of using the device to increase engagement in musical performances. Apart from perhaps feeling vibrations of the bass line, performances usually rely on sight and sound. Adding a tactile element could make for a more immersive experience, or help people with hearing impairment engage with the music. With the current tech, basic vibrating actuators can change the frequency of vibration to match the notes being played. While this can convey a simple melody, it lacks the richness of different instruments and musical components. The researchers’ full-freedom-of-motion actuator can convey a more vibrant sound. Voice, guitar, and drums, for instance, can each be converted into a delivery mechanism for a particular force. Like with vibration alone, the frequency of each force can be modulated to match the music. The experiment was exploratory, Rogers says, but it exploits the advanced capabilities of the system.

“Mooooo.” This dairy barn is full of cows, as you might expect. Cows are being milked, cows are being fed, cows are being cleaned up after, and a few very happy cows are even getting vigorously scratched behind the ears. “I wonder where the farmer is,” remarks my guide, Jan Jacobs. Jacobs doesn’t seem especially worried, though—the several hundred cows in this barn are being well cared for by a small fleet of fully autonomous robots, and the farmer might not be back for hours. The robots will let him know if anything goes wrong. At one of the milking robots, several cows are lined up, nose to tail, politely waiting their turn. The cows can get milked by robot whenever they like, which typically means more frequently than the twice a day at a traditional dairy farm. Not only is getting milked more often more comfortable for the cows, cows also produce about 10 percent more milk when the milking schedule is completely up to them. “There’s a direct correlation between stress and milk production,” Jacobs says. “Which is nice, because robots make cows happier and therefore, they give more milk, which helps us sell more robots.” Jan Jacobs is the human-robot interaction design lead for Lely, a maker of agricultural machinery. Founded in 1948 in Maassluis, Netherlands, Lely deployed its first Astronaut milking robot in the early 1990s. The company has since developed other robotic systems that assist with cleaning, feeding, and cow comfort, and the Astronaut milking robot is on its fifth generation. Lely is now focused entirely on robots for dairy farms, with around 135,000 of them deployed around the world. Essential Jobs on Dairy Farms The weather outside the barn is miserable. It’s late fall in the Netherlands, and a cold rain is gusting in from the sea, which is probably why the cows have quite sensibly decided to stay indoors and why the farmer is still nowhere to be found. Lely requires that dairy farmers who adopt its robots commit to letting their cows move freely between milking, feeding, and resting, as well as inside and outside the barn, at their own pace. “We believe that free cow traffic is a core part of the future of farming,” Jacobs says as we watch one cow stroll away from the milking robot while another takes its place. This is possible only when the farm operates on the cows’ schedule rather than a human’s. A conventional dairy farm relies heavily on human labor. Lely estimates that repetitive daily tasks represent about a third of the average workday of a dairy farmer. In the morning, the cows are milked for the first time. Most dairy cows must be milked at least twice a day or they’ll become uncomfortable, and so the herd will line up on their own. Traditional milking parlors are designed to maximize human milking efficiency. A milking carousel, for instance, slowly rotates cows as they’re milked so that the dairy worker doesn’t have to move between stalls. “We were spending 6 hours a day milking,” explains dairy farmer Josie Rozum, whose 120-cow herd at Takes Dairy Farm uses a pair of Astronaut A5 milking robots. “Now that the robots are handling all of that, we can focus more on animal care and comfort.”Lely An experienced human using well-optimized equipment can attach a milking machine to a cow in just 20 to 30 seconds. The actual milking takes only a few minutes, but with the average small dairy farm in North America providing a home for several hundred cows, milking typically represents a time commitment of 4 to 6 hours per day. There are other jobs that must be done every day at a dairy. Cows are happier with continuous access to food, which means feeding them several times a day. The feed is a mix of roughage (hay), silage (grass), and grain. The cows will eat all of this, but they prefer the grain, and so it’s common to see cows sorting their food by grabbing a mouthful and throwing it up into the air. The lighter roughage and silage flies farther than the grain does, leaving the cow with a pile of the tastier stuff as the rest gets tossed out of reach. This makes “feed pushing” necessary to shove the rest of the feed back within reach of the cow. And of course there’s manure. A dairy cow produces an average of 68 kilograms of manure a day. All that manure has to be collected and the barn floors regularly cleaned. Dairy Industry 4.0 The amount of labor needed to operate a dairy meant that until the early 1900s, most family farms could support only about eight cows. The introduction of the first milking machines, called bucket milkers, helped farmers milk 10 cows per hour instead of 4 by the mid-1920s. Rural electrification furthered dairy automation starting in the 1950s, and since then, both farm size and milk production have increased steadily. In the 1930s, a good dairy cow produced 3,600 kilograms of milk per year. Today, it’s almost 11,000 kilograms, and Lely believes that robots are what will enable small dairy farms to continue to scale sustainably. Lely But dairy robots are expensive. A milking robot can cost several hundred thousand dollars, plus an additional US $5,000 to $10,000 per year in operating costs. The Astronaut A5, Lely’s latest milking robot, uses a laser-guided robot arm to clean the cow’s udder before attaching teat cups one at a time. While the cow munches on treats, the Astronaut monitors her milk output, collecting data on 32 parameters, including indicators of the quality of the milk and the health of the cow. When milking is complete, the robot cleans the udder again, and the cow is free to leave as the robot steam cleans itself in preparation for the next cow. Lely argues that although the initial cost is higher than that of a traditional milking parlor, the robots pay for themselves over time through higher milk production (due primarily to increased milking frequency) and lower labor costs. Lely’s other robots can also save on labor. The Vector mobile robot handles continuous feeding and feed pushing, and the Discovery Collector is a robotic manure vacuum that keeps the floors clean. At Takes Dairy Farm, Rozum and her family used to spend several hours per day managing food for the cows. “The feeding robot is another amazing piece of the puzzle for our farm that allows us to focus on other things.”Takes Family Farm For most dairy farmers, though, making more money is not the main reason to get a robot, explains Marcia Endres, a professor in the department of animal science at the University of Minnesota. Endres specializes in dairy-cattle management, behavior, and welfare, and studies dairy robot adoption. “When we first started doing research on this about 12 years ago, most of the farms that were installing robots were smaller farms that did not want to hire employees,” Endres says. “They wanted to do the work just with family labor, but they also wanted to have more flexibility with their time. They wanted a better lifestyle.” Flexibility was key for the Takes family, who added Lely robots to their dairy farm in Ely, Iowa, four years ago. “When we had our old milking parlor, everything that we did as a family was always scheduled around milking,” says Josie Rozum, who manages the farm and a creamery along with her parents—Dan and Debbie Takes—and three brothers. “With the robots, we can prioritize our personal life a little bit more—we can spend time together on Christmas morning and know that the cows are still getting milked.” Takes Family Dairy Farm’s 120-cow herd is milked by a pair of Astronaut A5 robots, with a Vector and three Discovery Collectors for feeding and cleaning. “They’ve become a crucial part of the team,” explains Rozum. “It would be challenging for us to find outside help, and the robots keep things running smoothly.” The robots also add sustainability to small dairy farms, and not just in the short term. “Growing up on the farm, we experienced the hard work, and we saw what that commitment did to our parents,” Rozum explains. “It’s a very tough lifestyle. Having the robots take over a little bit of that has made dairy farming more appealing to our generation.” Takes Dairy Farm Of the 25,000 dairy farms in the United States, Endres estimates about 10 percent have robots. This is about a third of the adoption rate in Europe, where farms tend to be smaller, so the cost of implementing the robots is lower. Endres says that over the last five years, she’s seen a shift toward robot adoption at larger farms with over 500 cows, due primarily to labor shortages. “These larger dairies are having difficulty finding employees who want to milk cows—it’s a very tedious job. And the robot is always consistent. The farmers tell me, ‘My robot never calls in sick, and never shows up drunk.’ ” Endres is skeptical of Lely’s claim that its robots are responsible for increased milk production. “There is no research that proves that cows will be more productive just because of robots,” she says. It may be true that farms that add robots do see increased milk production, she adds, but it’s difficult to measure the direct effect that the robots have. “I have many dairies that I work with where they have both a robotic milking system and a conventional milking system, and if they are managing their cows well, there isn’t a lot of difference in milk production.” The Lely Luna cow brush helps to keep cows’ skin healthy. It’s also relaxing and enjoyable, so cows will brush themselves several times a day.Lely The robots do seem to improve the cows’ lives, however. “Welfare is not just productivity and health—it’s also the affective state, the ability to have a more natural life,” Endres says. “Again, it’s hard to measure, but I think that on most of these robot farms, their affective state is improved.” The cows’ relationship with humans changes too, comments Endres. When the cows no longer associate humans with being told where to go and what to do all the time, they’re much more relaxed and friendly toward people they meet. Rozum agrees. “We’ve noticed a tremendous change in our cows’ demeanor. They’re more calm and relaxed, just doing their thing in the barn. They’re much more comfortable when they can choose what to do.” Cows Versus Robots Cows are curious and clever animals, and have the same instinct that humans have when confronted with a new robot: They want to play with it. Because of this, Lely has had to cow-proof its robots, modifying their design and programming so that the machines can function autonomously around cows. Like many mobile robots, Lely’s dairy robots include contact-sensing bumpers that will pause the robot’s motion if it runs into something. On the Vector feeding robot, Lely product engineer René Beltman tells me, they had to add a software option to disable the bumper. “The cows learned that, ‘oh, if I just push the bumper, then the robot will stop and put down more feed in my area for me to eat.’ It was a free buffet. So you don’t want the cows to end up controlling the robot.” Emergency stop buttons had to be relocated so that they couldn’t be pressed by questing cow tongues. There’s also a social component to cow-robot interaction. Within their herd, cows have a well-established hierarchy, and the robots need to work within this hierarchy to do their jobs. For example, a cow won’t move out of the way if it thinks that another cow is lower in the hierarchy than it is, and it will treat a robot the same way. The engineers had to figure out how the Discovery Collector could drive back and forth to vacuum up manure without getting blocked by cows. “In our early tests, we’d use sensors to have the robot stop to avoid running into any of the cows,” explains Jacobs. “But that meant that the robot became the weakest one in the hierarchy, and it would just end up crying in the corner because the cows wouldn’t move for it. So now, it doesn’t stop.” One of the dirtiest jobs on a dairy farm is handled by the Discovery Collector, an autonomous manure vacuum. The robot relies on wheel odometry and ultrasonic sensors for navigation because it’s usually covered in manure.Evan Ackerman “We make the robot drive slower for the first week, when it’s being introduced to a new herd,” adds Beltman. “That gives the cows time to figure out that the robot is at the top of the hierarchy.” Besides maintaining their dominance at the top of the herd, the current generation of Lely robots doesn’t interact much with the cows, but that’s changing, Jacobs tells me. Right now, when a robot is driving through the barn, it makes a beeping sound to let the cows know it’s coming. Lely is looking into how to make these sounds more enjoyable for the cows. “This was a recent revelation for me,” Jacobs says. ”We’re not just designing interactions for humans. The cows are our users, too.” Human-Robot Interaction Last year, Jacobs and researchers from Delft University of Technology, in the Netherlands, presented a paper at the IEEE Human-Robot Interaction (HRI) Conference exploring this concept of robot behavior development on working dairy farms. The researchers visited robotic dairies, interviewed dairy farmers, and held workshops within Lely to establish a robot code of conduct—a guide that Lely’s designers and engineers use when considering how their robots should look, sound, and act, for the benefit of both humans and cows. On the engineering side, this includes practical things like colors and patterns for lights and different types of sounds so that information is communicated consistently across platforms. But there’s much more nuance to making a robot seem “reliable” or “friendly” to the end user, since such things are not only difficult to define but also difficult to implement in a way that’s appropriate for dairy farmers, who prioritize functionality. Jacobs doesn’t want his robots to try to be anyone’s friend—not the cow’s, and not the farmer’s. “The robot is an employee, and it should have a professional relationship,” he says. “So the robot might say ‘Hi,’ but it wouldn’t say, ‘How are you feeling today?’ ” What’s more important is that the robots are trustworthy. For Jacobs, instilling trust is simple: “You cannot gain trust by doing tricks. If your robot is reliable and predictable, people will trust it.” The electrically driven, pneumatically balanced robotic arm that the Lely Astronaut uses to milk cows is designed to withstand accidental (or intentional) kicks.Lely The real challenge, Jacobs explains, is that Lely is largely on its own when it comes to finding the best way of integrating its robots into the daily lives of people who may have never thought they’d have robot employees. “There’s not that much knowledge in the robot world about how to approach these problems,” Jacobs says. “We’re working with almost 20,000 farmers who have a bigger robot workforce than a human workforce. They’re robot managers. And I don’t know that there necessarily are other companies that have a customer base of normal people who have strategic dependence on robots for their livelihood. That is where we are now.” From Dairy Farmers to Robot Managers With the additional time and flexibility that the robots enable, some dairy farmers have been able to diversify. On our way back to Lely’s headquarters, we stop at Farm Het Lansingerland, owned by a Lely customer who has added a small restaurant and farm shop to his dairy. Large windows look into the barn so that restaurant patrons can watch the robots at work, caring for the cows that produce the cheese that’s on the menu. A self-guided tour takes you right up next to an Astronaut A5 milking robot, while signs on the floor warn of Vector feeding robots on the move. “This farmer couldn’t expand—this was as many cows as he’s allowed to have here,” Jacobs explains to me over cheese sandwiches. “So, he needs to have additional income streams. That’s why he started these other things. And the robots were essential for that.” The farmer is an early adopter—someone who’s excited about the technology and actively interested in the robots themselves. But most of Lely’s tens of thousands of customers just want a reliable robotic employee, not a science project. “We help the farmer to prepare not just the environment for the robots, but also the mind,” explains Jacobs. “It’s a complete shift in their way of working.” Besides managing the robots, the farmer must also learn to manage the massive amount of data that the robots generate about the cows. “The amount of data we get from the robots is a game changer,” says Rozum. “We can track milk production, health, and cow habits in real time. But it’s overwhelming. You could spend all day just sitting at the computer, looking at data and not get anything else done. It took us probably a year to really learn how to use it.” The most significant advantages to farmers come from using the data for long-term optimization, says the University of Minnesota’s Endres. “In a conventional barn, the cows are treated as a group,” she says. “But the robots are collecting data about individual animals, which lets us manage them as individuals.” By combining data from a milking robot and a feeding robot, for example, farmers can close the loop, correlating when and how the cows are fed with their milk production. Lely is doing its best to simplify this type of decision making, says Jacobs. “You need to understand what the data means, and then you need to present it to the farmer in an actionable way.” A Robotic Dairy All dairy farms are different, and farms that decide to give robots a try will often start with just one or two. A highly roboticized dairy barn might look something like this illustration, with a team of many different robots working together to keep the cows comfortable and happy. A: One Astronaut A5 robot can milk up to 60 cows. After the Astronaut cleans the teats, a laser sensor guides a robotic arm to attach the teat cups. Milking takes just a few minutes. B: In the feed kitchen, the Vector robot recharges itself while different ingredients are loaded into its hopper and mixed together. Mixtures can be customized for different groups of cows. C: The Vector robot dispenses freshly mixed food in small batches throughout the day. A laser measures the height of leftover food to make sure that the cows are getting the right amounts. D: The Discovery Collector is a mop and vacuum for cow manure. It navigates the barn autonomously and returns to its docking station to remove waste, refill water, and wirelessly recharge. E: As it milks, the Astronaut is collecting a huge amount of data—32 different parameters per teat. If it detects an issue, the farmer is notified, helping to catch health problems early. F: Automated gates control meadow access and will keep a cow inside if she’s due to be milked soon. Cows are identified using RFID collars, which also track their behavior and health. A Sensible Future for Dairy Robots After lunch, we stop by Lely headquarters, where bright red life-size cow statues guard the entrance and all of the conference rooms are dairy themed. We get comfortable in Butter, and I ask Jacobs and Beltman what the future holds for their dairy robots. In the near term, Lely is focused on making its existing robots more capable. Its latest feed-pushing robot is equipped with lidar and stereo cameras, which allow it to autonomously navigate around large farms without needing to follow a metal strip bolted to the ground. A new overhead camera system will leverage AI to recognize individual cows and track their behavior, while also providing farmers with an enormous new dataset that could allow Lely’s systems to help farmers make more nuanced decisions about cow welfare. The potential of AI is what Jacobs seems most excited about, although he’s cautious as well. “With AI, we’re suddenly going to take away an entirely different level of work. So, we’re thinking about doing research into the meaningfulness of work, to make sure that the things that we do with AI are the things that farmers want us to do with AI.” “The idea of AI is very intriguing,” comments Rozum. “I think AI could help to simplify things for farmers. It would be a tool, a resource. But we know our cows best, and a farmer’s judgment has to be there too. There’s just some component of dairy farming that you cannot take the human out of. Robots are not going to be successful on a farm unless you have good farmers.” Lely is aware of this and knows that its robots have to find the right balance between being helpful, and taking over. “We want to make sure not to take away the kinds of interactions that give dairy farmers joy in their work,” says Beltman. “Like feeding calves—every farmer likes to feed the calves.” Lely does sell an automated calf feeder that many dairy farmers buy, which illustrates the point: What’s the best way of designing robots to give humans the flexibility to do the work that they enjoy? “This is where robotics is going,” Jacobs tells me as he gives me a lift to the train station. “As a human, you could have two other humans and six robots, and that’s your company.” Many industries, he says, look to robots with the objective of minimizing human involvement as much as possible so that the robots can generate the maximum amount of value for whoever happens to be in charge. Dairy farms are different. Perhaps that’s because the person buying the robot is the person who most directly benefits from it. But I wonder if the concern over automation of jobs would be mitigated if more companies chose to emphasize the sustainability and joy of work equally with profit. Automation doesn’t have to be zero-sum—if implemented thoughtfully, perhaps robots can make work easier, more efficient, and more fun, too. Jacobs certainly thinks so. “That’s my utopia,” he says. “And we’re working in the right direction.”

In the three decades since Brewster Kahle spun up the nonprofit Internet Archive’s Wayback Machine, it has scaled up to include government websites and datasets—many of which are essential to the engineering and scientific communities. U.S. government agencies like the National Science Foundation, Department of Energy, and NASA are critical sources of research data, technical specifications, and standards documentation in pretty much every area where IEEE Spectrum’s audience works—AI & computer science, biomedical devices, power and energy, semiconductors, telecommunications…the list goes on. Access to that governmental data directly affects the reproducibility of experiments, the validation of models, and the integrity of the scholarly record. So what happens if an entire dataset vanishes? Among other things, it can invalidate years of research built upon that foundation. Until recently, wholesale deletion of data has been rare. In the United States, presidential transitions typically involve some changes to government websites to reflect new policy priorities. And after 9/11, the George W. Bush administration removed “millions of bytes” of information from government sites for security reasons as well as hundreds of Department of Defense documents and “tens of thousands” of Federal Energy Regulation Commission files. The Obama and Biden administrations likewise made changes to government websites but didn’t engage in large-scale removal of Web pages or datasets. Obama, in fact, expanded public access to government data in 2009 by launching Data.gov, whose stated mission is in part “to unleash the power of government open data to inform decisions by the public and policymakers.” During President Donald J. Trump’s first term, researchers at the Environmental Data & Governance Initiative found that some government sites became inaccessible, and the phrase “climate change” was purged from several government Web pages. Access to governmental data directly affects the reproducibility of experiments, the validation of models, and the integrity of the scholarly record. The second term has been different. In February, a few weeks after Trump was sworn in for his second term, The New York Times reported that his administration took down more than 8,000 Web pages and databases. Many of those pages have since reappeared, but some of the restored pages and files have had changes, including the erasure of terms like “climate change” (again) and “clean energy,” Grist reports. These moves have faced multiple court challenges; on 11 February, for instance, a federal judge ordered that public access to Web pages and datasets belonging to the Centers for Disease Control and Prevention and the Food and Drug Administration be restored. In our April issue, Spectrum Assistant Editor Gwendolyn Rak reports on efforts to preserve public access to information. In addition to the ongoing work at the Internet Archive, she describes how archivists at the Library Innovation Lab at Harvard Law School amassed a copy of the 16-terabyte archive of Data.gov, which includes more than 311,000 public datasets. That copied archive is being updated daily with new data hoovered up via automated queries to application programming interfaces (APIs). Archivists are the guardians of memory. We depend on them to help us stay in touch with our history, maintain our knowledge base, and provide context, allowing us to understand how we came to be where we are and to light the way forward. In the fields of science, engineering, and medicine, where today’s innovations stand on the shoulders of yesterday’s discoveries, these digital preservationists ensure that the circuit of human knowledge remains unbroken. This article appears in the April 2025 print issue as “Lots of Copies Keep Stuff Safe.” Editor’s note: This post was revised to match the print version.

This is a sponsored article brought to you by Freudenberg Sealing Technologies. The increasing deployment of collaborative robots (cobots) in outdoor environments presents significant engineering challenges, requiring highly advanced sealing solutions to ensure reliability and durability. Unlike industrial robots that operate in controlled indoor environments, outdoor cobots are exposed to extreme weather conditions that can compromise their mechanical integrity. Maintenance robots used in servicing wind turbines, for example, must endure intense temperature fluctuations, high humidity, prolonged UV radiation exposure, and powerful wind loads. Similarly, agricultural robots operate in harsh conditions where they are continuously exposed to abrasive dust, chemically aggressive fertilizers and pesticides, and mechanical stresses from rough terrains. To ensure these robotic systems maintain long-term functionality, sealing solutions must offer effective protection against environmental ingress, mechanical wear, corrosion, and chemical degradation. Outdoor robots must perform flawlessly in temperature ranges spanning from scorching heat to freezing cold while withstanding constant exposure to moisture, lubricants, solvents, and other contaminants. In addition, sealing systems must be resilient to continuous vibrations and mechanical shocks, which are inherent to robotic motion and can accelerate material fatigue over time. Comprehensive Technical Requirements for Robotic Sealing Solutions The development of sealing solutions for outdoor robotics demands an intricate balance of durability, flexibility, and resistance to wear. Robotic joints, particularly those in high-mobility systems, experience multidirectional movements within confined installation spaces, making the selection of appropriate sealing materials and geometries crucial. Traditional elastomeric O-rings, widely used in industrial applications, often fail under such extreme conditions. Exposure to high temperatures can cause thermal degradation, while continuous mechanical stress accelerates fatigue, leading to early seal failure. Chemical incompatibility with lubricants, fuels, and cleaning agents further contributes to material degradation, shortening operational lifespans. Friction-related wear is another critical concern, especially in robotic joints that operate at high speeds. Excessive friction not only generates heat but can also affect movement precision. In collaborative robotics, where robots work alongside humans, such inefficiencies pose safety risks by delaying response times and reducing motion accuracy. Additionally, prolonged exposure to UV radiation can cause conventional sealing materials to become brittle and crack, further compromising their performance. Advanced IPSR Technology: Tailored for Cobots To address these demanding conditions, Freudenberg Sealing Technologies has developed a specialized sealing solution: Ingress Protection Seals for Robots (IPSR). Unlike conventional seals that rely on metallic springs for mechanical support, the IPSR design features an innovative Z-shaped geometry that dynamically adapts to the axial and radial movements typical in robotic joints. Numerous seals are required in cobots and these are exposed to high speeds and forces.Freudenberg Sealing Technologies This unique structural design distributes mechanical loads more efficiently, significantly reducing friction and wear over time. While traditional spring-supported seals tend to degrade due to mechanical fatigue, the IPSR configuration eliminates this limitation, ensuring long-lasting performance. Additionally, the optimized contact pressure reduces frictional forces in robotic joints, thereby minimizing heat generation and extending component lifespans. This results in lower maintenance requirements, a crucial factor in applications where downtime can lead to significant operational disruptions. Optimized Through Advanced Simulation Techniques The development of IPSR technology relied extensively on Finite Element Analysis (FEA) simulations to optimize seal geometries, material selection, and surface textures before physical prototyping. These advanced computational techniques allowed engineers to predict and enhance seal behavior under real-world operational conditions. FEA simulations focused on key performance factors such as frictional forces, contact pressure distribution, deformation under load, and long-term fatigue resistance. By iteratively refining the design based on simulation data, Freudenberg engineers were able to develop a sealing solution that balances minimal friction with maximum durability. Furthermore, these simulations provided insights into how IPSR seals would perform under extreme conditions, including exposure to humidity, rapid temperature changes, and prolonged mechanical stress. This predictive approach enabled early detection of potential failure points, allowing for targeted improvements before mass production. By reducing the need for extensive physical testing, Freudenberg was able to accelerate the development cycle while ensuring high-performance reliability. Material Innovations: Superior Resistance and Longevity The effectiveness of a sealing solution is largely determined by its material composition. Freudenberg utilizes advanced elastomeric compounds, including Fluoroprene XP and EPDM, both selected for their exceptional chemical resistance, mechanical strength, and thermal stability. Fluoroprene XP, in particular, offers superior resistance to aggressive chemicals, including solvents, lubricants, fuels, and industrial cleaning agents. Additionally, its resilience against ozone and UV radiation makes it an ideal choice for outdoor applications where continuous exposure to sunlight could otherwise lead to material degradation. EPDM, on the other hand, provides outstanding flexibility at low temperatures and excellent aging resistance, making it suitable for applications that require long-term durability under fluctuating environmental conditions. To further enhance performance, Freudenberg applies specialized solid-film lubricant coatings to IPSR seals. These coatings significantly reduce friction and eliminate stick-slip effects, ensuring smooth robotic motion and precise movement control. This friction management not only improves energy efficiency but also enhances the overall responsiveness of robotic systems, an essential factor in high-precision automation. Extensive Validation Through Real-World Testing While advanced simulations provide critical insights into seal behavior, empirical testing remains essential for validating real-world performance. Freudenberg subjected IPSR seals to rigorous durability tests, including prolonged exposure to moisture, dust, temperature cycling, chemical immersion, and mechanical vibration. Throughout these tests, IPSR seals consistently achieved IP65 certification, demonstrating their ability to effectively prevent environmental contaminants from compromising robotic components. Real-world deployment in maintenance robotics for wind turbines and agricultural automation further confirmed their reliability, with extensive wear analysis showing significantly extended operational lifetimes compared to traditional sealing technologies. Safety Through Advanced Friction Management In collaborative robotics, sealing performance plays a direct role in operational safety. Excessive friction in robotic joints can delay emergency-stop responses and reduce motion precision, posing potential hazards in human-robot interaction. By incorporating low-friction coatings and optimized sealing geometries, Freudenberg ensures that robotic systems respond rapidly and accurately, enhancing workplace safety and efficiency. Tailored Sealing Solutions for Various Robotic Systems Freudenberg Sealing Technologies provides customized sealing solutions across a wide range of robotic applications, ensuring optimal performance in diverse environments. Automated Guided Vehicles (AGVs) operate in industrial settings where they are exposed to abrasive contaminants, mechanical vibrations, and chemical exposure. Freudenberg employs reinforced PTFE composites to enhance durability and protect internal components. Delta robots can perform complex movements at high speed. This requires seals that meet the high dynamic and acceleration requirements.Freudenberg Sealing Technologies Delta robots, commonly used in food processing, pharmaceuticals, and precision electronics, require FDA-compliant materials that withstand rigorous cleaning procedures such as Cleaning-In-Place (CIP) and Sterilization-In-Place (SIP). Freudenberg utilizes advanced fluoropolymers that maintain structural integrity under aggressive sanitation processes. Seals for Scara robots must have high chemical resistance, compressive strength and thermal resistance to function reliably in a variety of industrial environments.Freudenberg Sealing Technologies SCARA robots benefit from Freudenberg’s Modular Plastic Sealing Concept (MPSC), which integrates sealing, bearing support, and vibration damping within a compact, lightweight design. This innovation optimizes robot weight distribution and extends component service life. Six-axis robots used in automotive, aerospace, and electronics manufacturing require sealing solutions capable of withstanding high-speed operations, mechanical stress, and chemical exposure. Freudenberg’s Premium Sine Seal (PSS), featuring reinforced PTFE liners and specialized elastomer compounds, ensures maximum durability and minimal friction losses. Continuous Innovation for Future Robotic Applications Freudenberg Sealing Technologies remains at the forefront of innovation, continuously developing new materials, sealing designs, and validation methods to address evolving challenges in robotics. Through strategic customer collaborations, cutting-edge material science, and state-of-the-art simulation technologies, Freudenberg ensures that its sealing solutions provide unparalleled reliability, efficiency, and safety across all robotic platforms.

In the past few years, AI has set Silicon Valley on fire. The new book AI Valley: Microsoft, Google, and the Trillion-Dollar Race to Cash in on Artificial Intelligence chronicles those blazing high times, telling the stories of the startups, venture capital firms, and legacy tech companies that are burning bright—and those that have already flamed out. In the excerpt below, author Gary Rivlin tells the inside story of the startup Inflection, which was established in 2022 by LinkedIn founder Reid Hoffman and DeepMind founder Mustafa Suleyman. Inflection hoped to differentiate itself by building a chatbot with a high emotional intelligence, and the company was at one point valued at US $4 billion. But its chatbot, Pi, failed to gain market share and in March 2024 Microsoft acquired most of the company’s workforce, leaving what was left of Pi to be licensed for use as a foundation for customer service bots. Pi was not human and therefore could never have a personality. Yet it would fall on Inflection’s “personality team” to imbue Pi with a set of characteristics and traits that might make it seem like it did. The team’s ranks included several engineers, two linguists, and also Rachel Taylor, who had been the creative director of a London-based ad agency prior to going to work for Inflection. “Mustafa gave me a little bit of an overview of what they were working on, and I couldn’t stop thinking about it,” Taylor said. “I thought maybe it would be the most impactful thing I ever worked on.” Humans develop a personality through a complex interplay of genetics and environmental influences, including upbringing, culture, and life experiences. Pi’s personality began with the team listing traits. Some were positives. Be kind, be supportive. Others were negative traits to avoid, like irritability, arrogance, and combativeness. “You’re showing the model lots of comparisons that show it the difference between good and bad instances of that behavior,” Mustafa Suleyman said—“reinforcement learning with human feedback,” in industry parlance, or RLHF. Sometimes teams working on RLHF just label behavior they want a model to avoid (sexual, violent, homophobic). But Inflection had people assigning a numerical score to a machine’s responses. “That way the model basically learns, ‘Oh, this was a really good answer, I’m going to do more of that,’ or ‘That was terrible, I’m going to do less of that,’” said Anusha Balakrishnan, an Inflection engineer focused on fine-tuning. The scores were fed into an algorithm that adjusted the weighting of the model accordingly, and the process was repeated. Developing Pi’s Personality Traits Unlike many other AI companies, which outsourced reinforcement learning to third parties, Inflection hired and trained its own people. Applicants were put through a battery of tests, starting with a reading comprehension exercise that Suleyman described as “very nuanced and quite difficult.” Then came another set of exams and several rounds of training before they were put to work. The average “teacher” earned between $16 and $25 an hour, Suleyman said, but as much as $50 if someone was an expert in the right domain. “We try to make sure they come from a wide range of backgrounds and represent a wide range of ages,” Suleyman said. Inflection had many hundreds of teachers training Pi in the spring of 2023. “In some cases, we paid several hundred dollars an hour for very, very specialist people like behavioral therapists, psychologists, playwrights, and novelists,” Suleyman said. They even hired several comedians at one point, to help give Pi a sense of humor. “Our aim is a much more informal, relaxed, conversational experience,” Suleyman said. The company met a self-imposed deadline of March 12, 2023 for a beta version of Pi that they shared with thousands of testers. With its beta release, the company emerged from stealth mode. A press announcement described Pi as “a supportive and compassionate AI that is eager to talk about anything at any time.” The company described Pi a “new kind of AI” different than other chatbots on the market, By May, the app was free and available to anyone willing to register and sign in to use the service. The New York Times rarely runs even a short item about the release of a new product, especially one from a small, unknown startup. Yet few companies could boast of founders with the connections and star power of Inflection: Reid Hoffman, the co-founder of LinkedIn, and Suleyman, who was AI royalty as a cofounder of DeepMind. This clout translated into prime real estate on the front page of the Times Business section, including a large, eye-catching illustration and a headline that stretched across multiple columns: “My New BFF: Pi, an Emotional Support Chatbot.” Reporter Erin Griffith was skeptical of the breathing exercises that Pi suggested to help her relieve the stresses in her life. But the bot did help her develop a plan for managing a particularly hectic day, and it certainly left her feeling seen. Pi reassured Griffith that her feelings were “understandable,” “reasonable,” and “totally normal.” Suleyman posted a manifesto on the Inflection website on the day Pi was released. Social media basically had poisoned the world, he began. Outrage and anger drove engagement, and the lure of profits proved too strong. “Imagine an AI that helps you empathize with or even forgive ‘the other side,’ rather than be outraged by and fearful of them,” Suleyman wrote. “Imagine an AI that optimizes for your long-term goals and doesn’t take advantage of your need for distraction when you’re tired at the end of a long day.” He described the AI they were building as a “personal AI companion with the single mission of making you happier, healthier, and more productive.” In June 2023, Inflection announced its series A funding round. Suleyman and Hoffman had gone out thinking they would raise between $600 million and $675 million, but after the launch of Pi, Inflection was pegged as one of the hot new startups. A long list of investors wanted a piece. “We were overwhelmed with offers,” Suleyman said. In the end, they raised $1.3 billion on a venture round that valued Inflection at $4 billion. HarperCollins Publishers Inflection’s Technical and Business Challenges Pi’s willingness to tackle virtually any subject was a point of pride inside Inflection. Where other bots shut down users if they stepped anywhere near a sensitive topic, Pi invited a conversation. “It will try to acknowledge that a topic is sensitive or contentious and then be cautious about giving strong judgments and be led by the user,” Suleyman said. Pi corrected statements of fact that were wrong so as not to perpetuate misinformation but rather than outright reject a view, it offered counterevidence. Suleyman was particularly proud of Pi in the weeks after Hamas’s attack on Israel and the subsequent bombing campaign Israel waged in Gaza. “It was good in real time while things were unfolding, it’s good now,” he said two months into the hostilities. “It’s very balanced and evenhanded, very respectful.” If it had one bias, it was a deliberate one in favor of “peace and respect for human life,” Suleyman said. A bot that believed at its core in the sanctity of human life did not seem a bad thing. Taylor deemed the first version of Pi “acceptable.” “It was very, very polite and very formal,” she said. “But there wasn’t the conversationality we wanted.” Pleasant. Positive. Respectful. Those were all admirable traits but didn’t exactly add up to the “fun” experience they were selling. Yet finding that right balance proved difficult. The personality team would turn the dial up on one trait or another but it was as if they were playing Whac-A-Mole. They would fiddle with the weights and coax the model to use more slang and colloquialisms, but then Pi was “a little bit too friendly and informal in a way people might find rude,” Taylor said. The wide range of preferences among users was a consistent topic of conversation inside the company. Pi’s default mode was “friendly” but a short list of alternatives was added for people to choose from: casual, witty, compassionate, devoted. Pi would shift modes if a user told it they were looking for a sympathetic ear and not the friend who tries to fix a problem. But the future Pi, as imagined by Suleyman, was a model that read a person’s emotional tone and quickly adjusted on its own, much as someone might do if greeting a friend with a hearty hello but then switching immediately when learning they’re calling with bad news. But bots were not at the point where they could read a person’s preferences without clear instructions. It took at least ten turns of the conversation, Suleyman said, and as many as thirty to discern a user’s mood. “In the future, an AI is going to be many, many things all at once,” Suleyman said. “People ask me, ‘Is it a therapist?’ Well, it has flavors of therapist. It has flavors of a friend. It has flavors of supernerdy expert. It has flavors of coach and confidant.” Among their lofty goals was a Pi that had multiple personalities, like a cyborg Sybil with a dissociative identity disorder. As they saw it, Pi eventually would be able to assume a near-limitless number of modes able to match the moment. By December 2023, Pi was available for Android and its roughly 3 billion worldwide users. But Suleyman and others at Inflection were vague about user numbers—deliberately so. They were a disappointment. That fall, pollsters asked people who used chatbots which one they turned to most often. Fifty-two percent said ChatGPT and another 20 percent named Claude. Perplexity was third with a 10 percent share, followed by Google’s Bard (9 percent) and Bing (7 percent). Pi was lumped in with the 2 percent of users who selected “other.” The company had its usual long to-do list. Yet their main challenge was teaching Pi to get better at a wider range of tasks. People thought of Pi as a conversationalist, which was a good thing, but a helper that is good only at talking is limited. “Pi can’t code,” Balakrishnan said that winter. “It needs to get better at reasoning. It can’t take actions. It’s only really useful if you want to talk about your feelings.” From the book: AI Valley: Microsoft, Google, and the Trillion-Dollar Race to Cash In on Artificial Intelligence by Gary Rivlin. Copyright © 2025 by Gary Rivlin. Reprinted courtesy of Harper Business, an imprint of HarperCollins Publishers.

It’s been a busy three years since IEEE Women in Engineering celebrated its 25th anniversary in 2022. WIE facilitates the recruitment and retention of women in technical disciplines around the world. It also works to inspire girls to pursue an engineering career. There are student chapters at universities around the globe. Men may join the affinity group as well. Women make up less than a third of the world’s workforce in technology-related fields, according to a report by the World Bank Group. WIE has introduced several programs focused on increasing the number of women in the science, technology, engineering, and math fields. The new initiatives include grants and leadership programs, and several contests have been launched to encourage and support female students who are considering a STEM career. The group’s hard work is paying off. Winnie Ye, the current WIE chair and an electronics professor at Carleton University, which is just south of Ottawa, reports that, as of February, the number of members and student members was up over the same period last year. There are more than 28,800 members, a year-over-year increase greater than 17 percent. Student membership, at nearly 21,000, is up by roughly the same percentage. There are now more than 1,100 affinity groups. “Despite this growth, attracting and retaining paying members remains a challenge, particularly as students transition to professional membership,” Ye says. “Ensuring long-term engagement requires demonstrating clear value through career development, networking, and leadership opportunities.” A day of celebration, grants, and leadership training Many of the new programs were launched under Celia Shahnaz’s leadership. The WIE committee’s first elected chair, Shahnaz served in 2023 and 2024. In 2022 the committee made the chair an elected position. Shahnaz has been a WIE volunteer for more than 20 years. She established the first WIE group in the IEEE Bangladesh Section in 2010 and became its first chair. The IEEE senior member is a professor of electrical and electronic engineering at the Bangladesh University of Engineering and Technology, in Dhaka. She says she is most proud of bringing back the annual IEEE WIE Day, held on 29 June, the date the group was formally established in 1997. “The day is a member-engaging event targeting membership development and membership retention,” Shahnaz says. Activities include networking events, seminars, and workshops. This year’s theme is Pioneering Safe Cyberspace: Bridging Technology and Light for Security. For some groups, one day is not enough, so WIE Day events stretch out over several weeks. The inaugural event held in 2023 featured more than 120 activities worldwide. The number nearly doubled last year, with over 230 events. This year’s celebrations are scheduled to kick off on 23 June, to coincide with International Women in Engineering Day, then end on 7 July. The IEEE WIE Family Cares grant program was established to financially assist members with caregiving duties so they can attend an IEEE conference. A grant provides up to US $1,000 to cover expenses related to caring for children, senior citizens, and family members who are disabled. The grant is sponsored by the IEEE Foundation. To keep its members updated on industry trends, as well as research results and their practical applications, WIE has partnered with several IEEE societies and councils to offer a virtual Distinguished Lecturers program. More than 20 groups have provided speakers, including the Computer Society, Power Electronics Society, and Sensors Council. “I want to really strengthen WIE leadership with the local community so that we can provide more targeted resources and support women.” —Winnie Ye There’s also the new Industry Experts Network, a global database of industry professionals, researchers, and leaders who can be called upon to participate in an event. “For any of our events, we can access this list of amazing people, who we can ask to give talks and offer workshops and seminars,” Ye says. Several WIE programs are focused on providing women with leadership skills. Less than 10 percent of women hold positions such as CIO, CTO, and IT manager, or serve as technical team leaders, according to this year’s Women in Tech Stats report. The WIE International Leadership Summits, which are held around the world, provide opportunities for networking, mentorship, and collaboration. Eight were held last year, and seven are scheduled for this year in countries including Jordan, Pakistan, Poland, and the United States. The WIE Forum USA East helps its attendees develop and improve their leadership skills through talks by successful leaders. This year’s event is scheduled to be held from 6 to 8 November in Arlington, Va. Ye says she is most excited about the return of WIE’s International Leadership Conference. After a two-year absence, the event is scheduled to be held on 15 and 16 May in San Jose, Calif. The theme is Accelerating Leadership in an AI-Powered World. Registration is now open. STEM-related contests Several contests have been launched to encourage young women to pursue a STEM career, Shahnaz says. Women make up about 35 percent of STEM graduates—a gender disparity that hasn’t changed in the past 10 years, according to UNESCO’s 2024 Global Education Monitoring Report. Held for the first time in 2023, the WIE Climate Tech Big Idea Pitch competition encourages female engineering students and researchers to be entrepreneurial and is designed to increase the number of technical startups led by women. “I really wanted to inspire women to build their capacity using their technical knowledge and professional skills to be a startup founder,” Shahnaz says. “We wanted them to come up with a solution to climate-change-related problems and showcase their business ideas and models. We also wanted to nurture the talent of women in all membership grades including senior members, Fellows, and life Fellows.” In the 2023 contest, a team of engineering students from the Bangladesh University of Engineering and Technology won first place in the impact category for its design of a bamboo filter for diesel engines. There were 60 submissions that first year, and the number doubled last year. The deadline for submissions for this year’s contest will be announced soon. Capitalizing on the popularity of manga comics and graphic novels with young people, WIE launched a competition in 2023 to find the best-written manga that centered on Riko-chan, a fictional character who is a preuniversity student. Riko-chan uses STEM tools to help solve everyday problems. More than 40 people participated in the 2023 contest, and there were 81 submissions last year. Their stories are available to read online, and many have been translated into nine languages including French, Hindi, and Spanish. “We see this contest as an opportunity to showcase the diverse role models in engineering technology,” Ye says. “The goal is to spark curiosity among our younger audience and make STEM fields more relatable and more exciting for them.” This year’s manga competition is now accepting submissions. Check out the rules and deadlines on the WIE website. Mentoring and outreach programs A new mentoring program is in the works, Ye says. “We want to really create an active mentor-mentee matching and engagement platform within the WIE community,” she says. As part of her vision for a more engaged and inclusive community, she has launched the WIE Ambassador program. The initiative is designed to empower dedicated WIE members to advocate for IEEE’s mission globally. The ambassadors can promote WIE initiatives, organize local events, and inspire more women to pursue STEM careers, Ye says. She emphasizes the importance of expanding WIE’s presence in underrepresented regions such as China and South Africa. “During my term,” she says, “I’m committed to expanding our presence in these regions. I want to really strengthen WIE leadership with the local community so that we can provide more targeted resources and support women. We want to make sure that they are aware of us and become more active.”

I’m fascinated with the early 20th-century zeal for electrifying everyday things. Hand tools, toasters, hot combs—they all obviously benefited from the jolt of electrification. But the eraser? What was so problematic about the humble eraser that it needed electrifying? A number of things, it turned out. According to Hermann Lukowski in his 1935 patent application for an apparatus for erasing, “Hand held rubbers are clumsy and cover a greater area than may be required.” Aye, there’s the rub, as it were. Lukowski’s cone-tipped electric eraser, he argued, could better handle the fine detail. An electric eraser could also be a timesaver. In the days before computer-aided drawing and the ease of the delete and undo commands, manually erasing a document could be a delicate operation. Consider the careful technique Roscoe C. Sloane and John M. Montz suggest in their 1930 book Elements of Topographic Drawing. To make a correction to a map, these civil engineering professors at Ohio State University recommend the following steps: With a smooth, sharp knife pick the ink from the paper. This can be done without marring the surface. Place a hard, smooth surface, such as a [drafting] triangle, under the erasure before rubbing starts. When practically all the ink has been removed with the knife, rub with a pencil eraser. Erasing was not for the faint of heart! A Brief History of the Eraser Where did the eraser get its start? The British scientist Joseph Priestley is celebrated for his discovery of oxygen and not at all celebrated for his discovery of the eraser. Around 1766, while working on The History and Present State of Electricity, he found himself having to draw his own illustrations. First, though, he had to learn to draw, and because any new artist naturally makes mistakes, he also needed to erase. In 1766 or thereabouts, Joseph Priestley discovered the erasing properties of natural rubber.Alamy Alas, there weren’t a lot of great options for erasing at the time. For items drawn in ink, he could use a knife to scrape away errors; pumice or other rough stones could also be used to abrade the page and remove the ink. To erase pencil, the customary approach was to use a piece of bread or bread crumbs to gently grind the graphite off the page. All of the methods were problematic. Without extreme care, it was easy to damage the paper. Using bread was also messy, and as the writer and artist John Ruskin allegedly said, a waste of perfectly good bread. As the story goes, Priestley one day accidentally grabbed a piece of caoutchouc, or natural rubber, instead of bread and discovered that it could erase his mistakes. Priestley may have discovered this attribute of rubber, but Edward Nairne, an inventor, optician, and scientific-instrument maker, marketed it for sale. For three shillings (about one day’s wages for a skilled tradesman), you could purchase a half-inch (1.27-cm) cube of the material. Priestley acknowledged Nairne in the preface of his 1770 tutorial on how to draw, A Familiar Introduction to the Theory and Practice of Perspective, noting that caoutchouc was “excellently adapted to the purpose of wiping from paper the marks of a black-lead-pencil.” By the late 1770s, cubes of caoutchouc were generally known as rubbers or lead-eaters. Natural rubber might be good for erasing, but it isn’t necessarily an item you want sitting on your desk. It is extremely sensitive to temperature, becoming hard and brittle in the cold and soft and gummy in the heat. Over time, it inevitably degrades. And worst of all, it becomes stinky. What was so problematic about the humble eraser that it needed electrifying? Luckily, there were lots of other people looking for ways to improve natural rubber, and in 1839 Charles Goodyear developed the vulcanization process. By adding sulfur to natural rubber and then heating it, Goodyear discovered how to stabilize rubber in a firm state, what we would call today the thermosetting of polymers. In 1844 Goodyear patented a process to create rubber fabric. He went on to make rubber shoes and other products. (The tire company that bears his name was founded by the brothers Charles and Frank Seiberling several decades later.) Goodyear unfortunately died penniless, but we did get a better eraser out of his discovery. Who Really Invented the Electric Eraser? Albert Dremel, who opened his eponymous company in 1932, often gets credit for the invention of the electric eraser, but if that’s true, I can find no definitive proof. Out of more than 50 U.S. patents held by Dremel, none are for an electric eraser. In fact, other inventors may have a better claim, such as Homer G. Coy, who filed a patent for an electrified automatic eraser in 1927, or Ola S. Pugerud, who filed a patent for a rotatable electric eraser in 1906. The Dremel Moto-Tool, introduced in 1935, came with an array of swappable bits. One version could be used as an electric eraser.Dremel In 1935 Dremel did come out with the Moto-Tool, the world’s first handheld, high-speed rotary tool that had interchangeable bits for sanding, engraving, burnishing, and sharpening. One version of the Moto-Tool was sold as an electric eraser, although it was held more like a hammer than a pencil. Regardless of who invented the device, electric erasers were definitely on the market by 1929. Some of the earliest adopters were librarians, specifically those who maintained and had to frequently update the card catalog. Margaret Mann, an associate professor of library science at the University of Michigan, listed an electric eraser as recommended equipment in her 1930 book Introduction to Cataloging and the Classification of Books. She described a flat, round rubber eraser mounted on a motor-driven instrument similar to a dentist’s drill. The eraser could remove typewriting and print from catalog cards without leaving a rough appearance. By 1937, discussions of electric erasers were part of the library science curriculum at Columbia University. Electric erasers had gone mainstream. To erase pencil, the customary approach was to use a piece of bread to gently grind the graphite off the page. In 1930, the Charles Bruning Co.’s general catalog had six pages of erasers and accessories, with two pages devoted to the company’s electric erasing machine. Bruning, which specialized in engineering, drafting, and surveying supplies, also offered a variety of nonelectrified eraser products, including steel erasers (also known as desk knives), eraser shields (used to isolate the area to be erased), and a chisel-shaped eraser to put on the end of a pencil. The Loren Specialty Manufacturing Co. arrived late to the electric eraser game, introducing its first such product in 1953. Held in the hand like a pen or pencil, the Presto electric eraser would vibrate to abrade a small area in need of correction. The company spun off the Presto brand in 1962, about the time the Presto Model 80 [shown at top] was produced. This particular unit was used by officer workers at the New York Life Insurance Co. and is now housed at the Smithsonian’s Cooper Hewitt. The Creativity of the Eraser When I was growing up, my dad kept an electric eraser next to his drafting table. I loved playing with it, but it wasn’t until I began researching this article that I realized I had been using it all wrong. The pros know you’re supposed to shape the cylindrical rubber into a point in order to erase fine lines. Today almost all draftsmen, librarians, and office workers have gone digital, but some visual artists still use electric erasers. One of them is artist and educator Darrel Tank, who specializes in pencil drawings. I watched several of his surprisingly fascinating videos comparing various models of electric erasers. Seeing Tank use his favorite electric eraser to create texture on clothing or movement in hair made me realize that drawing is not just an additive process. Sometimes it is what’s removed that makes the difference. - YouTube Susan Piedmont-Palladino, an architect and professor at Virginia Tech’s Washington-Alexandria Architecture Center, has also thought a lot about erasing. She curated the exhibit “Tools of the Imagination: Drawing Tools and Technologies from the Eighteenth Century to the Present” at the National Building Museum in 2005 and authored the companion book of the same title. Piedmont-Palladino describes architectural design as a long process of doing, undoing, and redoing, deciding which ideas can stay and which must go. Piedmont-Palladino writes lovingly of a not-too-distant past, where this design process was captured in a building’s plans. When the architect was drafting by hand, the paper itself became a record of distress, showing where it had been scraped, erased, and redrawn. You could see points of uncertainty and points of decisiveness. But today, when almost all architectural drawing is done on a computer, users delete instead of erase. With a few keystrokes, an object can disappear, move to another spot, or miraculously reappear from the trash can. The design process is no longer etched into the page. Of course, the pencil, the eraser (electric or not), and the computer are all just tools for transmitting and visualizing ideas. The tools of any age reflect society in ways that aren’t always clear until new tools come to replace them. Both the pencil and the eraser had to be invented, and it is up to historians to make sure they aren’t forgotten. Part of a continuing series looking at historical artifacts that embrace the boundless potential of technology. An abridged version of this article appears in the April 2025 print issue as “When Electrification Came for the Eraser.” References The electric eraser, more than any object I have researched for Past Forward, has the most incorrect information about its history on the Internet—wrong names, bad dates, inaccurate assertions—which get repeated over and over again as fact. It’s a great reminder of the need to go back to original sources. As always, I enjoyed digging through patents to trace the history of invention and innovation in electric erasers. Other primary sources I consulted include Margaret Mann’s Introduction to Cataloging and the Classification of Books, a syllabus to Columbia University’s 1937 course on Library Service 201, and the Charles Bruning Co.’s 1930 catalog. Although Henry Petroski’s The Pencil: A History of Design and Circumstance only has a little bit of information on the history of erasers, it’s a great read about the implement that does the writing that needs to be erased.

Cheryl Conrad no longer seethes with the frustration that threatened to overwhelm her in 2006. As described in IEEE Spectrum, Cheryl’s husband, Tom, has a rare genetic disease that causes ammonia to accumulate in his blood. At an emergency room visit two decades ago, Cheryl told the doctors Tom needed an immediate dose of lactulose to avoid going into a coma, but they refused to medicate him until his primary doctor confirmed his medical condition hours later. Making the situation more vexing was that Tom had been treated at that facility for the same problem a few months earlier, and no one could locate his medical records. After Tom’s recovery, Cheryl vowed to always have immediate access to them. Today, Cheryl says, “Happily, I’m not involved anymore in lugging Tom’s medical records everywhere.” Tom’s two primary medical facilities use the same electronic health record (EHR) system, allowing doctors at both facilities to access his medical information quickly. In 2004, President George W. Bush set an ambitious goal for U.S. health care providers to transition to EHRs by 2014. Electronic health records, he declared, would transform health care by ensuring that a person’s complete medical information was available “at the time and place of care, no matter where it originates.” President George W. Bush looks at an electronic medical record system during a visit to the Cleveland Clinic on 27 January 2005. Brooks Kraft/Corbis/Getty Images Over the next four years, a bipartisan Congress approved more than US $150 million in funding aimed at setting up electronic health record demonstration projects and creating the administrative infrastructure needed. Then, in 2009, during efforts to mitigate the financial crisis, newly elected President Barack Obama signed the $787 billion economic stimulus bill. Part of it contained the Health Information Technology for Economic and Clinical Health Act, also known as the HITECH Act, which budgeted $49 billion to promote health information technology and EHRs in the United States. As a result, Tom, like most Americans, now has an electronic health record. However, many millions of Americans now have multiple electronic health records. On average, patients in the United States visit 19 different kinds of doctors throughout their lives. Further, many specialists have unique EHR systems that do not automatically communicate medical data between each other, so you must update your medical information for each one. Nevertheless, Tom now has immediate access to all his medical treatment and test information, something not readily available 20 years ago. Tom’s situation underlines the paradox of how far the United States has come since 2004 and how far it still must go to achieve President Bush’s vision of a complete, secure, easily accessible, and seamlessly interoperable lifetime EHR. As of 2021, nearly 80 percent of physicians and almost all nonfederal acute-care hospitals deployed an electronic health record system. For many patients in the United States today, instead of fragmented, paper medical record silos, they have a plethora of fragmented, electronic medical record silos. And thousands of health care providers are burdened with costly, poorly designed, and insecure EHR systems that have exacerbated clinician burnout, led to hundreds of millions of medical records lost in data breaches, and created new sources of medical errors. EHR’s baseline standardization does help centralize a very fragmented health care system, but in the rush to get EHR systems adopted, key technological and security challenges were overlooked and underappreciated. Subsequently, problems were introduced due to the sheer complexity of the systems being deployed. These still-unresolved issues are now potentially coupled with the unknown consequences of bolting on immature AI-driven technologies. Unless more thought and care are taken now in how to proceed as a fully integrated health care system, we could unintentionally put the entire U.S. health care system in a worse place than when President Bush first declared his EHR goal in 2004. IT to Correct Health Care Inefficiencies Is a Global Project Putting government pressure on the health care industry to adopt EHR systems through various financial incentives made sense by the early 2000s. Health care in the United States was in deep trouble. Spending increased from $74.1 billion in 1970 to more than $1.4 trillion by 2000, 2.3 times as fast as the U.S. gross domestic product. Health care costs grew at three times the rate of inflation from 1990 to 2000 alone, surpassing 13 percent of GDP. Two major studies conducted by the Institute of Medicine in 2000 and 2001, titled To Err Is Human and Crossing the Quality Chasm, found that health care was deteriorating in terms of accessibility, quality, and safety. Inferior quality and needless medical treatments, including overuse or duplication of diagnostic tests, underuse of effective medical practices, misuse of drug therapies, and poor communication between health care providers emerged as particularly frustrating problems. Administrative waste and unnecessary expenditures were substantial cost drivers, from billing to resolving insurance claims to managing patients’ cases. Health care’s administrative side was characterized as a “ monstrosity,” showing huge transaction costs associated with an estimated 30 billion communications conducted by mail, fax, or telephone annually at that time. Both health care experts and policymakers agreed that reductions in health care delivery and its costs were possible only by deploying health information technology such as electronic prescribing and EHR. Early adopters of EHR systems like the Mayo Clinic, Cleveland Clinic, and the U.S. Department of Veterans Affairs proved the case. Governments across the European Union and the United Kingdom reached the same conclusion. There has been a consistent push, especially in more economically advanced countries, to adopt EHR systems over the past two decades. For example, the E.U. has set a goal of providing 100 percent of its citizens across 27 countries access to electronic health records by 2030. Several countries are well on their way to this achievement, including Belgium, Denmark, Estonia, Lithuania, and Poland. Outside the E.U., countries such as Israel and Singapore also have very advanced systems, and after a rocky start, Australia’s My Health Record system seems to have found its footing. The United Kingdom was hoping to be a global leader in adopting interoperable health information systems, but a disastrous implementation of its National Programme for IT ended in 2011 after nine years and more than £10 billion. Canada, China, India, and Japan also have EHR system initiatives in place at varying levels of maturity. However, it will likely be years before they achieve the same capabilities found in leading digital-health countries. EHRs Need a Systems-Engineering Approach When it comes to embracing automation, the health care industry has historically moved at a snail’s pace, and when it does move, money goes to IT automation first. Market forces alone were unlikely to speed up EHR adoption. Even in the early 2000s, health care experts and government officials were confident that digitalization could reduce total health spending by 10 percent while improving patient care. In a highly influential 2005 study, the RAND Corp. estimated that adopting EHR systems in hospitals and physician offices would cost $98 billion and $17 billion, respectively. The report also estimated that these entities would save at least $77 billion a year after moving to digital records. A highly cited paper in HealthAffairs from 2005 also claimed that small physician practices could recoup their EHR system investments in 2.5 years and profit handsomely thereafter. Moreover, RAND claimed that a fully automated health care system could save the United States $346 billion per year. When Michael O. Leavitt, then the Secretary of Health and Human Services, looked at the projected savings, he saw them as “a key part of saving Medicare.” As baby boomers began retiring en masse in the early 2010s, cutting health care costs was also a political imperative since Medicare funding was projected to run out by 2020. Some doubted the EHR revolution’s health care improvement and cost reduction claims or that it could be achieved within 20 years. The Congressional Budget Office argued that the RAND report overstated the potential costs and benefits of EHR systems and ignored peer-reviewed studies that contradicted it. The CBO also pointed out that RAND assumed EHR systems would be widely adopted and effectively used, which implies that effective tools already existed, though very few commercially available systems were. There was also skepticism about whether replicating the benefits for early adopters of EHR systems—who spent decades perfecting their systems—was possible once the five-year period of governmental EHR adoption incentives ended. Even former House Speaker Newt Gingrich, a strong advocate for electronic health record systems, warned that health care was “30 times more difficult to fix than national defense.” The extent of the problem was one reason the 2005 National Academy of Sciences report, Building a Better Delivery System: A New Engineering / Health Care Partnership, forcefully and repeatedly called for innovative systems-engineering approaches to be developed and applied across the entire health care delivery process. The scale, complexity, and extremely short time frame for attempting to transform the totality of the health care environment demanded a robust “system of systems” engineering approach. This was especially true because of the potential human impacts of automation on health care professionals and patients. Researchers warned that ignoring the interplay of computer-mediated work and existing sociotechnical conditions in health care practices would result in unexpected, unintentional, and undesirable consequences. Additionally, without standard mechanisms for making EHR systems interoperable, many potential benefits would not materialize. As David Brailer, the first National Health Information Technology Coordinator, stated, “Unless interoperability is achieved…potential clinical and economic benefits won’t be realized, and we will not move closer to badly needed health care reform in the U.S.” HITECH’s Broken Promises and Unforeseen Consequences A few years later, policymakers in the Obama administration thought it was unrealistic to prioritize interoperability. They feared that defining interoperability standards too early would lock the health industry into outdated information-sharing approaches. Further, no existing health care business model supported interoperability, and a strong business model actively discouraged providers from sharing information. If patient information could easily shift to another provider, for example, what incentive does the provider have to readily share it? Instead, policymakers decided to have EHR systems adopted as widely and quickly as possible during the five years of HITECH incentives. Tackling interoperability would come later. The government’s unofficial operational mantra was that EHR systems needed to become operational before they could become interoperable. “Researchers have found that doctors spend between 3.5 and 6 hours a day (4.5 hours on average) filling out their digital health records.” Existing EHR system vendors, making $2 billion annually at the time, viewed the HITECH incentive program as a once-in-a-lifetime opportunity to increase market share and revenue streams. Like fresh chum to hungry sharks, the subsidy money attracted a host of new EHR technology entrants eager for a piece of the action. The resulting feeding frenzy pitted an IT-naïve health care industry rushing to adopt EHR systems against a horde of vendors willing to promise (almost) anything to make a sale. A few years into the HITECH program, a 2013 report by RAND wryly observed the market distortion caused by what amounted to an EHR adoption mandate: “We found that (EHR system) usability represents a relatively new, unique, and vexing challenge to physician professional satisfaction. Few other service industries are exposed to universal and substantial incentives to adopt such a specific, highly regulated form of technology, which has, as our findings suggest, not yet matured.” In addition to forcing health care providers to choose quickly among a host of immature EHR solutions, the HITECH program completely undercut the warnings raised about the need for systems engineering or considering the impact of automation on very human-centered aspects of health care delivery by professionals. Sadly, the lack of attention to these concerns affects current EHR systems. Today, studies like that conducted by Stanford Medicine indicate that nearly 70 percent of health care professionals express some level of satisfaction with their electronic health record system and that more than 60 percent think EHR systems have improved patient care. Electronic prescribing has also been seen as a general success, with the risk of medication errors and adverse drug events reduced. However, professional satisfaction with EHRs runs shallow. The poor usability of EHR systems surfaced early in the HITECH program and continues as a main driver for physician dissatisfaction. The Stanford Medicine study, for example, also reported that 54 percent of physicians polled felt their EHR systems detracted from their professional satisfaction, and 59 percent felt it required a complete overhaul. “What we’ve essentially done is created 24/7/365 access to clinicians with no economic model for that: The doctors don’t get paid.” —Robert Wachter, chair of the department of medicine at the University of California, San Francisco Poor EHR system usability results in laborious and low-value data entry, obstacles to face-to-face patient communication, and information overload, where clinicians have to wade through an excess of irrelevant data when treating a patient. A 2019 study in Mayo Clinic Proceedings comparing EHR system usability to other IT products like Google Search, Microsoft Word, and Amazon placed EHR products in the bottom 10 percent. Electronic health record systems were supposed to increase provider productivity, but for many clinicians, their EHRs are productivity vampires instead. Researchers have found that doctors spend between 3.5 and 6 hours a day (4.5 hours on average) filling out their patient’s digital health records, with an Annals of Internal Medicine study reporting that doctors in outpatient settings spend only 27 percent of their work time face-to-face with their patients. In those visits, patients often complain that their doctors spend too much time staring at their computers. They are not likely wrong, as nearly 70 percent of doctors in 2018 felt that EHRs took valuable time away from their patients. To address this issue, health care providers employ more than 100,000 medical scribes today—or about one for every 10 U.S. physicians—to record documentation during office visits, but this only highlights the unacceptable usability problem. Furthermore, physicians are spending more time dealing with their EHRs because the government, health care managers, and insurance companies are requesting more patient information regarding billing, quality measures, and compliance data. Patient notes are twice as long as they were 10 years ago. This is not surprising, as EHR systems so far have not complemented clinician work as much as directed it. “A phenomenon of the productivity vampire is that the goalposts get moved,” explains University of Michigan professor emeritus John Leslie King, who coined the phrase “productivity vampire.” King, a student of system–human interactions, continues, “With the ability to better track health care activities, more government and insurance companies are going to ask for that information in order for providers to get paid.” Robert Wachter, chair of the department of medicine at the University of California, San Francisco, and author of The Digital Doctor: Hope, Hype, and Harm at the Dawn of Medicine’s Computer Age, believes that EHRs “became an enabler of corporate control and outside entity control.” “It became a way that entities that cared about what the doctor was doing could now look to see in real time what the doctor was doing, and then influence what the doctor was doing and even constrain it,” Wachter says. Federal law mandates that patients have access to their medical information contained in EHR systems—which is great, says Wachter, but this also adds to clinician workloads, as patients now feel free to pepper their physicians with emails and messages about the information. “What we’ve essentially done is created 24/7/365 access to clinicians with no economic model for that: The doctors don’t get paid,” Wachter says. His doctors’ biggest complaints are that their EHR system has overloaded email inboxes with patient inquiries. Some doctors report that their in-boxes have become the equivalent of a second set of patients. It is not so much a problem with the electronic information system design per se, notes Wachter, but with EHR systems that “meet the payment system and the workflow system in ways that we really did not think about.” EHRs also promised to reduce stress among health care professionals. Numerous studies have found, however, that EHR systems worsen clinician burnout, with Stanford Medicine finding that 71 percent of physicians felt the systems contributed to burnout. Half of U.S. physicians are experiencing burnout, with 63 percent reporting at least one manifestation in 2022. The average physician works 53 hours weekly (19 hours more than the general population) and spends over 4 hours daily on documentation. Clinical burnout is lowest among clinicians with highly usable EHR systems or in specialties with the least interaction with their EHR systems, such as surgeons and radiologists. Physicians who make, on average, 4,000 EHR system clicks per shift, like emergency room doctors, report the highest levels of burnout. Aggravating the situation, notes Wachter, was “that decision support is so rudimentary…which means that the doctors feel like they’re spending all this time entering data in the machine, (but) getting relatively little useful intelligence out of it.” Poorly designed information systems can also compromise patient safety. Evidence suggests that EHR systems with unacceptable usability contribute to low-quality patient care and reduce the likelihood of catching medical errors. According to a study funded by the U.S. Agency for Healthcare Research and Quality, EHR system issues were involved in the majority of malpractice claims over a six-and-a-half-year period of study ending in 2021. Sadly, the situation has not changed today. Interoperability, Cybersecurity Bite Back EHR system interoperability closely follows poor EHR system usability as a driver of health care provider dissatisfaction. Recent data from the Assistant Secretary for Technology Policy / Office of the National Coordinator for Health Information Technology indicates that 70 percent of hospitals sometimes exchange patient data, though only 43 percent claim they regularly do. System-affiliated hospitals share the most information, while independent and small hospitals share the least. Exchanging information using the same EHR system helps. Wachter observes that interoperability among similar EHR systems is straightforward, but across different EHR systems, he says, “it is still relatively weak.” However, even if two hospitals use the same EHR vendor, communicating patient data can be difficult if each hospital’s system is customized. Studies indicate that patient mismatch rates can be as high as 50 percent, even in practices using the same EHR vendor. This often leads to duplicate patient records that lack vital patient information, which can result in avoidable patient injuries and deaths. The ability to share information associated with a unique patient identifier (UPI), like other countries that use advanced EHRs, including Estonia, Israel, and Singapore, makes health information interoperability easier, says Christina Grimes, digital health strategist for the Healthcare Information and Management Systems Society (HIMSS). But in the United States, “Congress has forbidden it since 1998” and steadfastly resists allowing for UPIs, she notes. Using a single-payer health insurance system, like most other countries with advanced EHR systems, would also make sharing patient information easier, decrease time spent on EHRs, and reduce clinician burnout, but that is also a nonstarter in the United States for the foreseeable future. Interoperability is even more challenging because an average hospital uses 10 different EHR vendors internally to support more than a dozen different health care functions, and an average health system has 16 different EHR vendors when affiliated providers are included. Grimes notes that only a small percentage of health care systems use fully integrated EHR systems that cover all functions. EHR systems adoption also promised to bend the national health care cost curve, but these costs continue to rise at the national level. The United States spent an estimated $4.8 trillion on health care in 2023, or 17.6 percent of GDP. While there seems to be general agreement that EHRs can help with cost savings, no rigorous quantitative studies at the national level show the tens of billions of dollars of promised savings that RAND loudly proclaimed in 2005. However, studies have shown that health care providers, especially those in rural areas, have had difficulty saving money by using EHR systems. A recent study, for example, points out that rural hospitals do not benefit as much from EHR systems as urban hospitals in terms of reducing operating costs. With 700 rural hospitals at risk of closing due to severe financial pressures, investing in EHR systems has not proved to be the financial panacea they thought it would be. Cybersecurity is a major cost not included in the 2005 RAND study. Even though there were warnings that cybersecurity was being given short shrift, vendors, providers, and policymakers paid scant attention to the cybersecurity implications of EHR systems, especially the multitude of new cyberthreat access points that would be created and potentially exploited. Tom Leary, senior vice president and head of government relations at HIMSS, points out the painfully obvious fact that “security was an afterthought. You have to make sure that security by design is involved from the beginning, so we’re still paying for the decision not to invest in security.” From 2009 to 2023, a total of 5,887 health care breaches of 500 records or more have been reported to the U.S. Department of Health and Human Services Office for Civil Rights resulting in some 520 million health care records being exposed. Health care breaches have also led to widespread disruption to medical care in various hospital systems, sometimes for over a month. In 2024, the average cost of a health care data breach was $9.97 million. The cost of these breaches will soon surpass the $27 billion ($44.5 billion in 2024 dollars) provided under HITECH to adopt EHRs. 2025 may see the first major revision since 2013 to the Health Insurance Portability and Accountability Act (HIPAA) Security Rule outlining how electronic protected health information will need to be cybersecured. The proposed rule will likely force health care providers and their EHR vendors to make cybersecurity investment a much higher priority. $100 Billion Spent on Health Care IT: Was the Juice Worth the (Mega) Squeeze? The U.S. health care industry has spent more than $100 billion on information technology, but few providers are fully meeting President Bush’s vision of a nation of seamlessly interoperable and secure digital health records. Many past government policymakers now admit they failed to understand the complex business dynamics, technical scale, complexity, or time needed to create a nationwide system of usable, interoperable EHR systems. The entire process lacked systems-engineering thinking. As Seema Verma, former administrator of the Centers for Medicare and Medicaid Services, told Fortune, “We didn’t think about how all these systems connect with one another. That was the real missing piece.” Over the past eight years, successive administrations and congresses have taken actions to try to rectify these early oversights. In 2016, the 21st Century Cures Act was passed, which kept EHR system vendors and providers from blocking the sharing of patient data, and spurred them to start working in earnest to create a trusted health information exchange. The Cures Act mandated standardized application programming interfaces (APIs) to promote interoperability. In 2022, the Trusted Exchange Framework and Common Agreement (TEFCA) was published, which aims to facilitate technical principles for securely exchanging health information. “The EHR venture has proved troublesome thus far. The trouble is far from over.” —John Leslie King, University of Michigan professor emeritus In late 2023, the first Qualified Health Information Networks (QHINs) were approved to begin supporting the exchange of data governed by TEFCA, and in 2024, updates were made to the APIs to make information interoperability easier. These seven QHINs allow thousands of health providers to more easily exchange information. Combined with the emerging consolidation among hospital systems around three EHR vendors—Epic Systems Corp., Oracle Health, and Meditech—this should improve interoperability in the next decade. These changes, says HIMSS’s Tom Leary, will help give “all patients access to their data in whatever format they want with limited barriers. The health care environment is starting to become patient-centric now. So, as a patient, I should soon be able to go out to any of my healthcare providers to really get that information.” HIMSS’s Christina Grimes adds that the patient-centric change is the continuing consolidation of EHR system portals. “Patients really want one portal to interact with instead of the number they have today,” she says. In 2024, the Assistant Secretary for Technology Policy / Office of the National Coordinator for Health IT, the U.S. government department responsible for overseeing electronic health systems’ adoption and standards, was reorganized to focus more on cybersecurity and advanced technology like AI. In addition to the proposed HIPAA security requirements, Congress is also considering new laws to mandate better cybersecurity. There is hope that AI can help overcome EHR system usability issues, especially clinician burnout and interoperability issues like patient matching. Wachter states that the new AI scribes are showing real promise. “The way it works is that I can now have a conversation with my patient and look the patient in the eye. I’m actually focusing on them and not my keyboard. And then a note, formatted correctly, just magically appears. Almost ironically, this new set of AI technologies may well solve some of the problems that the last technology created.” Whether these technologies live up to the hype remains to be seen. More concerning is whether AI will exacerbate the rampant feeling among providers that they have become tools of their tools and not masters of them. As EHR systems become more usable, interoperable, and patient-friendly, the underlying foundations of medical care can be finally addressed. High-quality evidence backs only about 10 percent of the care patients receive today. One of the great potentials of digitizing health records is to discover what treatments work best and why and then distribute that information to the health care community. While this is an active research area, more research and funding are needed. Twenty years ago, Tom Conrad, who himself was a senior computer scientist, told me he was skeptical that having more information necessarily meant that better medical decisions would automatically be made. He pointed out that when doctors’ earnings are related to the number of patients they see, there is a trade-off between the better care that EHR provides and the sheer amount of time required to review a more complete medical record. Today, the trade-off is not in the patients’ or doctors’ favor. Whether it can ever be balanced is one of the great unknowns. Obviously, no one wants to go back to paper records. However, as John Leslie King says, “The way forward involves multiple moving targets due to advances in technology, care, and administration. Most EHR vendors are moving as fast as they can.” However, it would be foolish to think it will be smooth sailing from here on, King says: “The EHR venture has proved troublesome thus far. The trouble is far from over.”

In Mira Daher’s home country of Lebanon, the national grid provides power for only a few hours a day. The country’s state-owned energy provider, Electricity of Lebanon (EDL), has long struggled to meet demand, and a crippling economic crisis that began in 2019 has worsened the situation. Most residents now rely on privately owned diesel-powered generators for the bulk of their energy needs. But in recent years, the rapidly falling cost of solar panels has given Lebanese businesses and families a compelling alternative, and the country has seen a boom in private solar-power installations. Total installed solar capacity jumped nearly eightfold between 2020 and 2022 to more than 870 megawatts, primarily as a result of off-grid rooftop installations. Daher, head of tendering at the renewable-energy company Earth Technologies, in Antelias, Lebanon, has played an important part in this ongoing revolution. She is in charge of bidding for new solar projects, drawing up designs, and ensuring that they are correctly implemented on-site. “I enjoy the variety and the challenge of managing diverse projects, each with its own unique requirements and technical hurdles,” she says. “And knowing that my efforts also contribute to a sustainable future for Lebanon fills me with pride and motivates me a lot.” An Early Mentor Daher grew up in the southern Lebanese city of Saida (also called Sidon) where her father worked as an electrical engineer in the construction sector. His work helped to inspire her interest in technology at a young age, she says. When she was applying for university, he encouraged her to study electrical engineering too. “My first mentor was my father,” says Daher. “He increased my curiosity and passion for technology and engineering, and when I watched him work and solve complex problems, that motivated me to follow in his footsteps.” In 2016, she enrolled at Beirut Arab University to study electrical and electronics engineering. When she graduated in 2019, Daher says, the country’s solar boom was just taking off, which prompted her to pursue a master’s degree in power and energy, with a specialization in solar power, at the American University of Beirut. “My thesis concentrated on the energy situation in Lebanon and explored potential solutions to increase the reliance on renewable resources,” she says. “Five or six years ago, solar systems had high costs. But today the cost [has] decreased a lot because of new technologies, and because there is a lot of production of solar panels in China.” Entering the Workforce After graduating in 2021, Daher started a job as a solar-energy engineer at the Beirut-based solar-power company Mashriq Energy, where she was responsible for developing designs and bids for new solar installations, similar to her current role. It was a steep learning curve, Daher says, because she had to quickly pick up business skills, including financial modeling and contract negotiations. She also learned to deal with the practicalities of building large solar developments, such as site constraints and regulations. In 2022, she joined Earth Technologies as a solar project design engineer. Various organizations, including Lebanese government and nongovernmental agencies such as the United Nations, request bids for solar power installations they want to build—a process known as tendering. Daher’s principal responsibility is to prepare and submit bids for these projects, but she also supervises their implementation. Daher’s role requires her to maintain a broad base of knowledge about the solar projects she oversees.Mira Daher “I oversee the entire project cycle, from identifying and managing tenders to designing, pricing, and implementing solar projects across residential, industrial, commercial, and utility sectors,” she says. The first step in the process is to visit the proposed installation site to determine where solar panels should be positioned based on the landscape and local weather conditions. Once this is done, Daher and her team come up with a design for the plant. This involves figuring out what kinds of solar panels, inverters, and batteries will fit the budget and how to wire all the components together. The team runs simulations of the proposed plant to ensure that the design meets the client’s needs. Daher is then responsible for negotiating with the client to make sure that the proposal fulfills their technical and budgetary requirements. Once the client has approved the design, other teams oversee construction of the plant, though Daher says she makes occasional site visits to ensure the design is being implemented correctly. Daher’s role requires her to have a solid understanding of all the components that go into a solar plant, from the different brands of power electronics to the civil engineering required to build supporting structures for solar panels. “You have to know everything about the project,” she says. Solar Power for Development Earth Technologies operates across the Middle East and Africa, but Daher says most of the solar installations she works on are in Lebanon. Some of the most interesting have been development-focused projects funded by the U.N. Daher led the U.N.-funded installation of solar panels at nine hospitals, as well as a project that uses solar power to pump water to people in remote parts of the country. More recently, she has started work on a solar and battery installation for street lighting in the town of Bourj Hammoud, which will allow shops to stay open later and help to boost the local economy. The projects she has overseen generally cost around US $700,000 to $800,000. But securing funding for renewable projects is an ongoing challenge in Lebanon, says Daher, given the uncertain economic situation. More recently, the country was also rocked by the conflict between Israel and the Lebanon-based paramilitary group Hezbollah. This resulted in widespread bombing of Beirut, the capital, and the country’s southern regions last October and November. “The two months of conflict were incredibly challenging,” says Daher. “The environment was unsafe and filled with uncertainty, leaving us constantly anxious about what the future held.” Safety concerns forced her to relocate from her home in Beirut to a village called Ain El Jdideh. This meant she had to drive about an hour and a half on unsafe roads to get to work. Several of the major projects she was working on were also halted as they were in the areas that bore the brunt of the conflict. One U.N.-funded project she worked on in Ansar, in southern Lebanon, was knocked offline when an adjacent building was destroyed. “Despite these hardships, we persevered, and I am grateful that the war has ended, allowing us to regain some stability and security,” says Daher. A Challenging But Fulfilling Career Despite these difficulties, Daher remains optimistic about the future of renewable energy in Lebanon, and she says it can be a deeply rewarding career. Breaking into the industry requires a strong educational foundation, though, so she recommends first pursuing a degree focused on power systems and renewable technologies. The energy sector is a male-dominated field, says Daher, which can make it difficult for women to find their footing. “I’ve often encountered biases, stereotypes that can make it more difficult to be taken seriously, or to have my voice heard,” she adds. “Overcoming these obstacles requires resilience, confidence, and a commitment to demonstrating my expertise and capabilities.” It also requires a commitment to continual learning, due to the continued advances being made in solar-power technology. “It’s very important to stay up to date,” she says. “This field is always evolving. Every day, you can see a lot of new technologies.”

Your weekly selection of awesome robot videos Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE Spectrum robotics. We also post a weekly calendar of upcoming robotics events for the next few months. Please send us your events for inclusion. RoboSoft 2025: 23–26 April 2025, LAUSANNE, SWITZERLAND ICUAS 2025: 14–17 May 2025, CHARLOTTE, NC ICRA 2025: 19–23 May 2025, ATLANTA, GA London Humanoids Summit: 29–30 May 2025, LONDON IEEE RCAR 2025: 1–6 June 2025, TOYAMA, JAPAN 2025 Energy Drone & Robotics Summit: 16–18 June 2025, HOUSTON, TX RSS 2025: 21–25 June 2025, LOS ANGELES ETH Robotics Summer School: 21–27 June 2025, GENEVA IAS 2025: 30 June–4 July 2025, GENOA, ITALY ICRES 2025: 3–4 July 2025, PORTO, PORTUGAL IEEE World Haptics: 8–11 July 2025, SUWON, KOREA IFAC Symposium on Robotics: 15–18 July 2025, PARIS RoboCup 2025: 15–21 July 2025, BAHIA, BRAZIL RO-MAN 2025: 25–29 August 2025, EINDHOVEN, NETHERLANDS Enjoy today’s videos! This robot can walk, without electronics, and only with the addition of a cartridge of compressed gas, right off the 3D-printer. It can also be printed in one go, from one material. Researchers from the University of California San Diego and BASF, describe how they developed the robot in an advanced online publication in the journal Advanced Intelligent Systems. They used the simplest technology available: a desktop 3D-printer and an off-the-shelf printing material. This design approach is not only robust, it is also cheap—each robot costs about $20 to manufacture. And details! [ Paper ] via [ University of California San Diego ] Why do you want a humanoid robot to walk like a human? So that it doesn’t look weird, I guess, but it’s hard to imagine that a system that doesn’t have the same arrangement of joints and muscles that we do will move optimally by just trying to mimic us. [ Figure ] I don’t know how it manages it, but this little soft robotic worm somehow moves with an incredible amount of personality. Soft actuators are critical for enabling soft robots, medical devices, and haptic systems. Many soft actuators, however, require power to hold a configuration and rely on hard circuitry for control, limiting their potential applications. In this work, the first soft electromagnetic system is demonstrated for externally-controlled bistable actuation or self-regulated astable oscillation. [ Paper ] via [ Georgia Tech ] Thanks, Ellen! A 180-degree pelvis rotation would put the “break” in “breakdancing” if this were a human doing it. [ Boston Dynamics ] My colleagues were impressed by this cooking robot, but that may be because journalists are always impressed by free food. [ Posha ] This is our latest work about a hybrid aerial-terrestrial quadruped robot called SPIDAR, which shows unique and complex locomotion styles in both aerial and terrestrial domains including thrust-assisted crawling motion. This work has been presented in the International Symposium of Robotics Research (ISRR) 2024. [ Paper ] via [ Dragon Lab ] Thanks, Moju! This fresh, newly captured video from Unitree’s testing grounds showcases the breakneck speed of humanoid intelligence advancement. Every day brings something thrilling! [ Unitree ] There should be more robots that you can ride around on. [ AgileX Robotics ] There should be more robots that wear hats at work. [ Ugo ] iRobot, who pioneered giant docks for robot vacuums, is now moving away from giant docks for robot vacuums. [ iRobot ] There’s a famous experiment where if you put a dead fish in current, it starts swimming, just because of its biomechanical design. Somehow, you can do the same thing with an unactuated quadruped robot on a treadmill. [ Delft University of Technology ] Mush! Narrowly! [ Hybrid Robotics ] It’s freaking me out a little bit that this couple is apparently wandering around a huge mall that is populated only by robots and zero other humans. [ MagicLab ] I’m trying, I really am, but the yellow is just not working for me. [ Kepler ] By having Stretch take on the physically demanding task of unloading trailers stacked floor to ceiling with boxes, Gap Inc has reduced injuries, lowered turnover, and watched employees get excited about automation intended to keep them safe. [ Boston Dynamics ] Since arriving at Mars in 2012, NASA’s Curiosity rover has been ingesting samples of Martian rock, soil, and air to better understand the past and present habitability of the Red Planet. Of particular interest to its search are organic molecules: the building blocks of life. Now, Curiosity’s onboard chemistry lab has detected long-chain hydrocarbons in a mudstone called “Cumberland,” the largest organics yet discovered on Mars. [ NASA ] This University of Toronto Robotics Institute Seminar is from Sergey Levine at UC Berkeley, on Robotics Foundation Models. General-purpose pretrained models have transformed natural language processing, computer vision, and other fields. In principle, such approaches should be ideal in robotics: since gathering large amounts of data for any given robotic platform and application is likely to be difficult, general pretrained models that provide broad capabilities present an ideal recipe to enable robotic learning at scale for real-world applications.From the perspective of general AI research, such approaches also offer a promising and intriguing approach to some of the grandest AI challenges: if large-scale training on embodied experience can provide diverse physical capabilities, this would shed light not only on the practical questions around designing broadly capable robots, but the foundations of situated problem-solving, physical understanding, and decision making. However, realizing this potential requires handling a number of challenging obstacles. What data shall we use to train robotic foundation models? What will be the training objective? How should alignment or post-training be done? In this talk, I will discuss how we can approach some of these challenges. [ University of Toronto ]

The U.S. government recommends that everyone have a disaster kit that includes a weather radio. These radios tune to a nationwide network run by the National Oceanic and Atmospheric Administration (NOAA) and the Federal Communications Commission that provides alerts about hazardous weather and other major emergencies. Such broadcasts can be a lifeline when other communication systems go out. But what if you could step it up and get not just audio information but also images, charts, and written reports, even while completely off the grid? Turns out you can, thanks to modern geosynchronous weather satellites, and it’s never been easier than with KrakenRF’s new Discovery Dish system. This system involves buying a US $115 70-centimeter-diameter parabolic antenna, and then one of a number of $109 swappable feeds that cover different frequency bands. To try out the system, I got one feed suitable for picking up L-band satellite transmissions, and another tuned for detecting the radio emissions from galactic hydrogen clouds. The parabolic antenna comes as three metal petals plus some ancillary bits and pieces for holding the feed and mounting the dish on a support. Everything is held together with nuts and bolts, so it can be dissembled and reassembled, and the petals are light and stack together nicely—you could pack them in a suitcase if you ever wanted to travel and sample a different sky. In addition to KrakenRF’s dish and feed, you’ll also need a software-defined radio (SDR) receiver and a computer with software to decode the signals coming from the feed. Many SDRs can be used, but you’ll need one that comes with what’s known as a bias tee built in, or you’ll need to add a bias tee yourself. The bias tee supplies power to the low-noise amplifiers used in KrakenRF’s feeds. I used the recommended $34 RTL-SDR Blog V3 (which comes as a USB dongle), with my MacBook, but you can use a PC or Raspberry Pi as a host computer as well. The Discovery Dish is formed by three petals [top center] that bolt together with other mounting gear [top left and right]. Feeds are mounted on a pole and adjusted to be level with the dish’s focus [bottom]. Different feeds allow different applications, such as 1680 megahertz for receiving L-band satellite transmissions or 1420 MHz for radio astronomy. A software-defined radio receiver decodes signals.James Provost After I inserted the L-band feed into the dish, it was time to look for a satellite. Following KrakenRF’s guide, I used Carl Reinemann’s Web app to print out a list of azimuths and elevations for geosynchronous weather satellites based on my location. Then I headed up to the roof of my New York City apartment building with the mast from my portable ham radio antenna to provide a mount. And then I headed straight back down again when I realized that it was too blustery for a temporary mount. The dish is perforated with holes to reduce air resistance, but there was still a real risk of the wind toppling the portable mast and sweeping it over the side of the building. A couple of days later, I returned to calmer conditions, and with my iPhone employed as a compass and inclinometer, I pointed the dish at the coordinates for the GOES-East weather satellite. This satellite hangs over the equator at a longitude of 75 degrees west, close to that of New York City. A second satellite, GOES-West, sits at 135 degrees west, over the Pacific Ocean. These GOES satellites are fourth-generation spacecraft in a long line of invaluable weather satellites that have been operated by NOAA and the U.S. National Weather Service for 50 years. The first of the current generation, known as GOES-R, launched in 2016 and features a number of upgrades. For radio enthusiasts, the most significant of the upgrades are its downlink broadcast capabilities. The current GOES satellites transmit images taken in multiple wavelengths and scales. A false-color full-disk image [above] is captured in an infrared band that detects moisture and ash; the image at top shows the eastern United States in an approximation of what you would see with the naked eye. Stephen Cass/NOAA The GOES-R satellites transmit data at 400 kilobits per second, versus a maximum of 128 Kb/s for previous generations, allowing more information to be included, such as images from other weather satellites around the globe. The satellites also merge satellite-image data and emergency and weather information into a single link that can be simultaneously picked up by one receiver, instead of needing two as previously. For fine dish-pointing adjustments, I was guided by watching the signal in the frequency spectrum analyzer built into SatDump, an open-software package designed for decoding satellite transmissions picked up by SDR receivers. I groaned when no matter how I adjusted the dish, I could barely get the signal above the noise. But much to my surprise, I nonetheless started seeing an image of the Earth begin to form on the display. The original GOES-R design specified that receiving ground dishes would have to be at least one meter in diameter, but the folks at KrakenRF have built their feeds around a new ultralow-noise amplifier that can make the weaker signal gathered by their smaller dish usable. Soon I had pictures of the Earth in multiple wavelengths, both raw and in false color, with and without the superimposed outlines of states and countries, plus a wide assortment of other charts plotting rainfall and wind speeds for different areas. The GOES satellites also broadcast information uploaded from the U.S. National Weather Service, such as this chart of marine wind speeds.Stephen Cass/National Weather Service My next test was to do a spot of radio astronomy, swapping out the L-band feed for the galactic hydrogen emission feed. Getting results was a much longer process: First I had to point the dish at a bit of the sky where I knew the Milky Way wasn’t to obtain baseline data (done with the help of the Stellarium astronomy site). Then I pointed the dish straight up and waited for the rotation of the Earth to bring the Milky Way into view. Pulling the signal out of the noise is a slow process—you have to integrate 5 minutes of data from the receiver—but eventually a nice curve formed that indicated I was still safely within the embrace of the spiral arms of our home galaxy. Much more sophisticated radio astronomy can be done, especially if you mount the dish on a scanning platform to generate 2D maps. But I swapped back the L-band feed just to marvel at how our planet looks from 36,000 kilometers away!

Marko Delimar has been a proponent of empowering the next generation of engineers, scientists, and technologists since he was an undergraduate engineering student at the University of Zagreb, in Croatia. The IEEE senior member now mentors undergraduate and graduate students at his alma mater, where he is a professor of electrical engineering and computing. IEEE has played a key role in his quest to provide students with the support they need, he says. Marko Delimar Employer: University of Zagreb, in Croatia Title: Professor of electrical engineering and computing Member grade: Senior member Alma mater: University of Zagreb Throughout his 30 years of volunteering, Delimar has worked to build a community for students. He founded the University of Zagreb’s IEEE student branch and later became its chair. He went on to become the branch’s counselor and a member of the IEEE Croatia Section’s student activities committee. He has held numerous IEEE leadership positions, and he served on the organization’s Board of Directors. To engage more student members and help connect student branches worldwide, he helped found the IEEEXtreme programming competition, an annual, 24-hour virtual contest in which teams compete to solve computer coding problems. He is continuing his mission as the 2025 president of the IEEE Foundation, focusing on how the organization’s charitable partner can help students and young professionals prosper. Thanks to donations, the Foundation is able to fund scholarships, research and travel grants, and fellowships in partnership with IEEE societies and sections. “Supporting IEEE programs is something that I’m very proud of,” Delimar says. His goals as president, he says, are to increase awareness of the Foundation’s donor-supported programs and to persuade more people to support its causes. Supporting the next generation of engineers After learning about IEEE from several of his professors who were members, Delimar joined the organization in 1994 during his second year at the University of Zagreb. But without a student branch at the school, there was no local community for student members. That year he successfully petitioned IEEE to establish the first student branch in Croatia. He served as its chair until he graduated with his EE bachelor’s degree in 1996. Through his involvement, he was simultaneously “learning about the organization and volunteering,” he says, adding that it helped him better understand IEEE. After graduating, he joined his alma mater as a teaching assistant and researcher. He was later hired as a faculty member. He also conducted research in power engineering under his former professor, IEEE Senior Member Zdravko Hebel, who is known for his work on the Croatian power transmission network. Delimar continued to volunteer, serving as chair of student activities for the IEEE Croatia Section until 2001. He also was the counselor for the student branch. “For me, IEEE Foundation Day highlights how the IEEE Foundation is more than a charitable organization—it is the heart of IEEE’s philanthropic efforts, where generosity meets impact.” Within four years of his guidance, “the branch was collaborating with other branches in not only Region 8 [Europe, Middle East, and Africa] but also around the world,” he says. He decided it was time to spread his wings, and he began volunteering for the region. His first position was as the 2005–2006 vice chair of the IEEE Region 8 student activities committee, which is responsible for student programs and benefits. At the time, IEEE was having trouble retaining student members, he says. “Students would graduate and not renew their membership,” he says. “There was also an issue with some of the student branches, as they were not communicating well or collaborating with each other.” Delimar and IEEE Member Ricardo Varela, who was also on the committee, brainstormed how to better engage students and increase their participation. The two men wanted to create an event that would allow students across the world to participate at the same time. “It sounded like a very crazy idea,” he says, “because it’s nighttime for one half of the world and daytime for the other half. You can’t even hold a meeting at the same time everywhere, let alone an activity.” To overcome the time-zone issue, Delimar and Varela devised a 24-hour competition on programming, which was popular among engineering students at the time, Delimar says. Having the contest take place over 24 hours ensured all participants were on equal footing, he says. Forty teams participated in the first IEEE Xtreme competition, which was held in October 2006. It has since grown in popularity. Last year nearly 8,800 teams from 75 countries participated. Although he’s not involved in the contest anymore, Delimar says he’s proud of its success. In 2007 he became vice chair of the IEEE Region 8 membership activities committee, which plans events for members. He was then elected as the 2010–2011 Region 8 director, and in 2013 he became IEEE secretary. Both are Board-level positions. “Being a part of the IEEE Board of Directors gave me the opportunity to learn about and serve on several interesting committees that were trying to reach particular goals at the time, such as increasing member engagement, improving training for new IEEE officers, and refining IEEE’s ability to quickly adapt to the fast-changing environment,” Delimar says. His time on the board inspired him to advocate for the formation of an ad hoc committee on European public policy activities. He served as its chair, and in 2018 it became a permanent committee. Renamed the IEEE European Public Policy Committee, it supports members of the European Union and European Free Trade Association in developing technology-related policies. Delimar was its chair until 2020. “IEEE has been able to provide a united, unbiased voice of what is good for technology and what is good for Europe,” he says. “It has been very well received by the European Commission.” In 2016 Pedro Ray, the 2010 IEEE president, asked Delimar to be a volunteer for the IEEE Foundation, and he joined the board the next year. “It’s been a very rewarding experience,” he says. Leading the IEEE Foundation Delimar says that his main goal as president is to increase awareness among IEEE members of the Foundation and its programs. “The Foundation supports more than 250 funds and programs, and I want to strengthen its connections and partnerships across IEEE,” he says. To accomplish that goal, the Foundation has been raising its visibility. In 2023 it celebrated its 50th anniversary with a reception in New York City. Other celebratory activities were held that year at the IEEE Vision, Innovation, and Challenges Summit and Honors Ceremony and the IEEE Power & Energy Society General Meeting. Last year the Foundation established 16 February as IEEE Foundation Day. The annual celebration marks the day in 1973 when the philanthropic organization was launched. The inaugural event was designed to reflect the Foundation’s vision of being the heart of IEEE’s charitable giving. This year’s celebration focused on students and young professionals, highlighting beneficiaries of scholarships, grants, and fellowships and the impact they have had on the recipients. “For me, IEEE Foundation Day highlights how the IEEE Foundation is more than a charitable organization—it is the heart of IEEE’s philanthropic efforts, where generosity meets impact,” Delimar says. “Our donor-supported programs—like scholarships, travel grants, awards, research grants, and competitions—are more than financial support for our students and young professionals; they are catalysts for making dreams come true.” He says he wants to engage members who aren’t typically donors and thus expand the Foundation’s reach. “I want to enable people with different professional journeys, economic backgrounds, cultures, and geography to be able to participate as donors for the IEEE Foundation,” he says. “Every donor—whether they are a student, young professional, or IEEE life member—is important.” Visit the IEEE Foundation website to discover upcoming events, learn ways to make a gift, and see how the organization’s charitable efforts are making an impact.

IEEE Spectrum is rebooting our careers newsletter! In partnership with tech career development company Taro, every issue will be bringing you deeper insight into how to pursue your goals and navigate professional challenges. Sign up now to get insider tips, expert advice, and practical strategies delivered to your inbox for free. One of my close friends is a hiring manager at Google. She recently posted about an open position on her team and was immediately overwhelmed with applications. We’re talking about thousands of applicants within days. What surprised me most, however, was the horrendous quality of the average submission. Most applicants were obviously unqualified or had concocted entirely fake profiles. The use of generative AI to automatically fill out (and, in some cases, even submit) applications is harmful to everyone; employers are unable to filter through the noise, and legitimate candidates have a harder time getting noticed—much less advancing to an interview. This problem exists even for companies that don’t have the magnetism of the Google brand. Recruiting is a numbers game with slim odds. As AI becomes increasingly mainstream, the job search can feel downright impossible. So how can job seekers stand out among the deluge of candidates? When there are hundreds or thousands of applicants, the best way to distinguish yourself is by leveraging your network. With AI, anyone with a computer can trivially apply to thousands of jobs. On the other hand, people are restricted by Dunbar’s number—the idea that humans can maintain stable social relationships with only about 150 people. Being one of those 150 people is harder, but it also carries more weight than a soulless job application. A referral from a trusted connection immediately elevates you as a promising candidate. Your goal, therefore, is to aggressively pursue opportunities based on people you’ve worked with. A strong referral has two benefits: Your profile gets increased visibility. You “borrow” the credibility of the person who gave a vote of confidence. So how do you get one of these coveted referrals? Start with groups you’re part of that have a well-defined admission criteria. Most commonly, this will be your university or workplace. Engage with university alumni working in interesting roles, or reconnect with an ex-colleague to see what they’re up to. Good luck out there!—Rahul ICYMI: Despite 2024 Layoffs, Tech Jobs Expected to Take Off In 2024, the technology sector saw additional cuts after massive layoffs in 2022 and 2023. Despite these numbers, however, engineers seem to be doing just fine. U.S. employment for electrical engineers is expected to grow 9 percent from 2023 to 2033, compared with 4 percent for all occupations. And the World Economic Forum’s Future of Jobs Report 2025, revealed that technology-related roles are the fastest-growing categories globally, with the most rapid growth in demand anticipated for big data specialists, financial-technology engineers, AI specialists, and software developers. Read more at https://spectrum.ieee.org/tech-jobs ICYMI: Electric Vehicles Made These Engineers Expendable When veteran Wall Street Journal reporter Mike Colias began writing about the automotive industry in 2010, the internal-combustion engine still served as the beating heart of legacy carmakers. Since then, the hard pivot to electric vehicles has sidelined engine design and upended a century of internal order at these companies. Colias has observed the transformation, and the recent detour back to plug-in hybrids, from a front-row seat in Detroit. Read an excerpt from his new book Inevitable: Inside the Messy, Unstoppable Transition to Electric Vehicles, which tells the tale of one power-train engineer at Ford whose internal-combustion-engine expertise slowly became expendable.

In my career, I’ve often been the only woman in a room full of men, a situation all too common in tech-related fields. From the start of my engineering journey, I was among a handful of women in my graduate program. This trend continued when I began my first software engineering job as the only woman on my team, and eight years later, I remain a minority in the tech world. Women currently constitute about 35 percent of the technical workforce. This statistic highlights the ongoing challenge of gender disparity in tech, where women have historically been underrepresented. And the gap becomes even more pronounced in leadership roles, with women making up only about 25 percent of CEOs in the technology sector, according to one report. While the gender gap remains, many women have successful careers in tech. With these five actionable tips, women can take charge and own their space in the engineering world. Negotiate Your Compensation with Confidence Women engineers in the United States earn about 10 percent less than their male counterparts in similar roles. When I started my career, I believed that as long as I worked on solving the problems I enjoyed, my compensation didn’t really matter. But over time, I realized this mind-set is flawed. Compensation isn’t just about money. It reflects your value and contributions to a company. Salary negotiation can feel daunting, especially if you feel uncertain about what’s fair or struggle to articulate your worth. But this is your chance to own your value. The first step in negotiating your salary is learning your market value. Websites like Glassdoor and Levels.fyi provide salary insights based on data from employees in similar roles. They’re excellent starting points to get a sense of base salaries, bonuses, stock options, and other compensation factors. Your network can provide another valuable source of information. Ask trusted friends or colleagues working at similar companies what someone with your experience would typically be paid. Phrasing the question as a hypothetical avoids putting anyone on the spot to disclose their salary, while still giving you helpful information. Once you clearly understand your market worth, you are in a solid position to negotiate your compensation, whether switching jobs or discussing a raise with your current employer. If you are in the midst of a job switch and have received an offer, you’ve already proven your value to the new company. The hiring team has invested significant time and resources in the interview process and is eager to bring you onboard. They’re often willing to negotiate at this point, but it’s essential to know when you have leverage. One of the best times to negotiate is when you have more than one offer or are approaching the final rounds of interviews with other companies. This creates a healthy competition for your skills. There are also ways to prepare if you’re approaching a compensation cycle in your current company. Start early: Begin conversations with your boss three to five months before you expect changes, as compensation is often finalized months ahead. You should also know your company’s policies; some employers may increase salary, while others focus on benefits or long-term career growth. Compensation isn’t just salary—stock and bonuses also matter. When you’re ready, talk to your manager confidently and with the right data. A good manager will appreciate your asking for what you deserve. Postdoctoral researcher Caitlin McCowan adjusts a customized scanning tunneling microscope. Craig Fritz Don’t Attribute Your Success to Gender When organizations are committed to advancing diversity, studies suggest that the public tends to perceive women’s promotions as driven less by their intelligence and effort, and more by their gender. If those perceptions also prevail within their company, women might be told they have an unfair advantage. These types of remarks can make it easy to start questioning, for example, whether you have truly earned a promotion. If you’re encountering such attitudes, you can instead embrace your accomplishments and take ownership of the work that led to them. Keeping a “brag list” to record your achievements and strengths can serve as a reminder of your true capabilities. This list can include concrete results or milestones you’ve reached through effort and skill, as well as personal qualities that contribute to your success. Creating a brag list isn’t about feeding your ego. It’s about reminding yourself of the hard work you’ve put in and acknowledging that you belong because of your talents. Nandu Koripally [front] and Lulu Yao work with a structural supercapacitor developed by engineers at the University of California, San Diego.David Baillot/UC San Diego Jacobs School of Engineering Open Doors That Are Closed to You If a door isn’t open to you, it doesn’t mean you don’t belong on the other side. I’m often surprised by how much women miss out on simply because we don’t ask about the opportunities that interest us. If you think you’re capable of an opportunity, clearly express interest to your manager. When you approach them, explain the type of opportunity you seek, such as leading a project or transitioning to a new role. Then highlight your strengths by connecting your request to your previous successes. This shows that you’ve delivered in the past and are ready for the next step. Identifying where you can add value will also make your request more compelling. If you see a gap or area for improvement, frame your request around how you can help the team or organization. If the opportunity isn’t immediately available, ask for feedback on your readiness and how to prepare for future opportunities. This shows you’re eager to learn and improve. Follow up regularly to express your continued interest as well. Paula Kirya, a mechanical engineering graduate student at the University of California, San Diego, studies light-manipulating micro- and nanostructures on Morpho butterfly wings to assess the level of fibrosis in cancer biopsy samples.David Baillot/UC San Diego Jacobs School of Engineering Practice Authentic Leadership As women, we often possess leadership traits that differ from what’s expected of men. These are shaped by societal expectations, cultural influences, and workplace dynamics that historically defined leadership in a particular way. The traits that are often associated with women can be viewed negatively in leadership due to gender biases. I’ve worked with many highly empathetic women who would make excellent leaders. However, they may be perceived as weak because traditional leadership norms prioritize assertiveness and authority over emotional intelligence. Traits like fostering relationships shouldn’t be seen as weaknesses simply because they don’t fit traditional leadership molds. Rather, these qualities can bring a different approach to leadership. It can be helpful to create a Venn diagram highlighting your strengths, improvement areas, and the overlap between them. This process may reveal characteristics that aren’t necessarily flaws but, when harnessed effectively, can become strengths. By recognizing these traits and being intentional in their application, you can transform them into key advantages in your leadership style. You don’t need to conform to a specific image of what leadership should look like. Instead, practice authentic leadership by providing guidance in a way that’s true to you. See Yourself as a Leader My final piece of advice is a simple one: Leadership isn’t just for other people—it’s for you, too. When my manager first asked if I would like to take on a new role as the team lead, I was thrilled. But when I went home, self-doubt and anxiety clouded my excitement. The image I had of a technical leader was that of a man, and I couldn’t envision myself in that role at first. Over time, however, I changed that mental image. Start visualizing yourself as a leader in your organization, regardless of your current position. Leadership is less about title and more about mind-set. Take initiative, lead by example, and make decisions that contribute to your team’s success. When you start thinking and acting like a leader, others will also begin to see you that way. Navigating the tech industry as a woman can be challenging, but it’s important to recognize and embrace your value. By confidently negotiating compensation, attributing your success to your skills, asking about new opportunities, embracing authentic leadership, and seeing yourself as a leader, you can carve out your space in the engineering world. These strategies will not only empower you, but contribute to a more inclusive and diverse tech industry.

EPICS in IEEE, an Educational Activities program, celebrated its 15th anniversary last year. The Engineering Projects in Community Service initiative provides nonprofit organizations with technology to improve and deliver services to their community while broadening undergraduate EE students’ hands-on experience with engineering-related topics. The program reached new heights last year by distributing more than US $226,000 to 39 projects, engaging more than 900 students and 1,400 volunteers and IEEE members including proposal reviewers, mentors, and project support participants. The standout year caps Stephanie Gillespie’s three-year term as chair of the EPICS in IEEE committee. An IEEE member, she is an associate dean at the University of New Haven engineering college, in Connecticut. Under her leadership, the EPICS program streamlined processes for collecting data and boosted storytelling efforts to illustrate the impact of its activities. Proposal applications nearly doubled, and approved projects increased by 44 percent. “During my three years as committee chair, I’ve loved seeing so many students worldwide have the opportunity to engage with their community and truly consider the needs of their project stakeholders,” Gillespie says. “There are so many ways in which those positive experiences can trickle through their professional and personal lives. “New leadership will bring new ideas, and I’m excited for the future of EPICS in IEEE and the continued growth of this outstanding program.” Pedro Wightman is this year’s committee chair. He is an IEEE senior member and an associate professor in the engineering school at the Universidad del Rosario, in Bogotá. Wightman, who has been an active committee member, helped lead an EPICS project himself. “EPICS in IEEE represents a bridge between engineering theory and community needs,” Wightman says. “It brings these two realities together for humanitarian purposes and helps polish an engineering student’s academic experience so that they’re not only technology experts but humans first. “I hope the program continues to grow and strengthen, because there’s so much need out there but also so many young engineers around the world willing to put their energy and knowledge toward improving the well-being of people.” The following projects, all funded last year, showcase the enthusiasm and dedication of students to create and deploy solutions to help people. Augmented reality system for distance learning A lack of access to trained teachers and quality educational programs threatens students’ development, especially in rural areas with poor infrastructure and limited resources. A team from the Mehran University of Engineering and Technology (MUET) in Jamshoro, Pakistan, addressed such challenges with its Augmented Reality 3D System for Interactive Learning in Rural Elementary Schools project. It reached hundreds of pupils in underserved areas with a virtual, sustainable, and hands-on learning platform that engages students and teachers alike. The team received a$4,115 grant to design and deploy their prototype. “Many rural areas lack access to qualified educators and modern teaching resources,” says Sameer Qazi, an IEEE student member and a MUET senior. “Our project seeks to bridge that gap by providing an interactive and immersive learning system that delivers an innovative educational experience with a focus on science, art, and history.” The team from Mehran University of Engineering and Technology, in Jamshoro, Pakistan, assembles its augmented-reality prototype of its distance learning system. Mehran University of Engineering and Technology EPICS in IEEE project team The project involved nine students and two faculty supervisors from the university’s electronic engineering department. Each member contributed to one of the three project modules: content development, power management, and the display system. Qazi was project lead for the display system module. The team partnered with the Fast Rural Development Program, an organization dedicated to transforming underprivileged areas and driving sustainable development in rural parts of Pakistan’s Sindh province. The partnership helped ensure the team’s solution was aligned with community needs and could be more seamlessly integrated into the target schools’ curriculum. “I hope the program continues to grow and strengthen, because there’s so much need out there but also so many young engineers around the world willing to put their energy and knowledge toward improving the well-being of people.” —Pedro Wightman “From funding to mentorship, EPICS in IEEE provided us with essential support that allowed us to bring this project from concept to reality,” Qazi says. “I encourage anyone considering an EPICS project to seize the opportunity. It’s a chance to make a tangible impact on society while growing as an engineer and leader.” Solar-powered greenhouse A creative team of students from BridgeValley Community and Technical College in South Charleston, W.Va., used its $3,200 grant to establish their Solar PV–Powered Smart Sustainable Greenhouse project. It is designed to optimize plant growth by maintaining ideal conditions and conserving energy through remote monitoring. The team used Arduino and other technologies, and it partnered with Café Appalachia, an eco-friendly nonprofit with its own farms. The group developed a greenhouse system featuring solar-powered lithium-ion batteries, smart irrigation, fan control, and IoT integration for real-time data monitoring. Other components included tracking charge controllers, soil-moisture sensors, solenoid valves, flow meters, temperature and humidity sensors, a 356-millimeter fan, and a Wi-Fi module with data logging capabilities. “We chose to work with Café Appalachia because they had an interest in sustainable greenhouse practices and offered space for the system,” says team lead Joshua “Youngil” Kim, an IEEE member. “Their dedication to giving back to their community and addressing climate change aligned with our goals.” Kim, a former BridgeValley educator, now is an assistant professor of electrical engineering at Charleston Southern University, in South Carolina. The team expressed gratitude for the funding from the IEEE Antennas and Propagation Society, an EPICS in IEEE partner. “EPICS’ support of this greenhouse project has been instrumental in allowing us to contribute to our community and offer educational opportunities to our students,” Kim says. “Thanks to EPICS in IEEE’s generosity and commitment to fostering educational growth, it’s been a valuable experience for everyone involved.” To learn more about EPICS projects or to submit a project proposal, join the mailing list and receive monthly updates.

A long-awaited, emerging computer network component may finally be having its moment. At Nvidia’s GTC event last week in San Jose, the company announced that it will produce an optical network switch designed to drastically cut the power consumption of AI data centers. The system—called a co-packaged optics, or CPO, switch—can route tens of terabits per second from computers in one rack to computers in another. At the same time, startup Micas Networks, announced that it is in volume production with a CPO switch based on Broadcom’s technology. In data centers today, network switches in a rack of computers consist of specialized chips electrically linked to optical transceivers that plug into the system. (Connections within a rack are electrical, but several startups hope to change this.) The pluggable transceivers combine lasers, optical circuits, digital signal processors, and other electronics. They make an electrical link to the switch and translate data between electronic bits on the switch side and photons that fly through the data center along optical fibers. Co-packaged optics is an effort to boost bandwidth and reduce power consumption by moving the optical/electrical data conversion as close as possible to the switch chip. This simplifies the setup and saves power by reducing the number of separate components needed and the distance electronic signals must travel. Advanced packaging technology allows chipmakers to surround the network chip with several silicon optical-transceiver chiplets. Optical fibers attach directly to the package. So all the components are integrated into a single package except for the lasers, which remain external because they are made using nonsilicon materials and technologies. (Even so, CPOs require only one laser for every eight data links in Nvidia’s hardware.) “An AI supercomputer with 400,000 GPUs is actually a 24-megawatt laser.” —Ian Buck, Nvidia As attractive a technology as that seems, its economics have kept it from deployment. “We’ve been waiting for CPO forever,” says Clint Schow, a co-packaged optics expert and IEEE Fellow at the University of California, Santa Barbara, who has been researching the technology for 20 years. Speaking of Nvidia’s endorsement of technology, he said the company “wouldn’t do it unless the time was here when [GPU-heavy data centers] can’t afford to spend the power.” The engineering involved is so complex, Schow doesn’t think it’s worthwhile unless “doing things the old way is broken.” And indeed, Nvidia pointed to power consumption in upcoming AI data centers as a motivation. Pluggable optics consume “a staggering 10 percent of the total GPU compute power” in an AI data center, says Ian Buck, Nvidia’s vice president of hyperscale and high-performance computing. In a 400,000-GPU factory, that would translate to 40 megawatts, and more than half of that goes just to powering the lasers in a pluggable optics transceiver. “An AI supercomputer with 400,000 GPUs is actually a 24-megawatt laser,” he says. Optical Modulators One fundamental difference between Broadcom’s scheme and Nvidia’s is the optical modulator technology that encodes electronic bits onto beams of light. In silicon photonics there are two main types of modulators—Mach-Zehnder, which Broadcom uses and is the basis for pluggable optics, and microring resonator, which Nvidia chose. In the former, light traveling through a waveguide is split into two parallel arms. Each arm can then be modulated by an applied electric field, which changes the phase of the light passing through. The arms then rejoin to form a single waveguide. Depending on whether the two signals are now in phase or out of phase, they will cancel each other out or combine. And so electronic bits can be encoded onto the light. Microring modulators are far more compact. Instead of splitting the light along two parallel paths, a ring-shaped waveguide hangs off the side of the light’s main path. If the light is of a wavelength that can form a standing wave in the ring, it will be siphoned off, filtering that wavelength out of the main waveguide. Exactly which wavelength resonates with the ring depends on the structure’s refractive index, which can be electronically manipulated. However, the microring’s compactness comes with a cost. Microring modulators are sensitive to temperature, so each one requires a built-in heating circuit, which must be carefully controlled and consumes power. On the other hand, Mach-Zehnder devices are considerably larger, leading to more lost light and some design issues, says Schow. That Nvidia managed to commercialize a microring-based silicon photonics engine is “an amazing engineering feat,” says Schow. Nvidia CPO Switches According to Nvidia, adopting the CPO switches in a new AI data center would lead to one-fourth the number of lasers, boost power efficiency for trafficking data 3.5-fold, improve the on-time reliability of signals traveling from one computer to another by 63 times, make networks tenfold more resilient to disruptions, and allow customers to deploy new data-center hardware 30 percent faster. “By integrating silicon photonics directly into switches, Nvidia is shattering the old limitation of hyperscale and enterprise networks and opening the gate to million-GPU AI factories,” said Nvidia CEO Jensen Huang. - YouTube youtu.be The company plans two classes of switch, Spectrum-X and Quantum-X. Quantum-X, which the company says will be available later this year, is based on InfiniBand network technology, a network scheme more oriented to high-performance computing. It delivers 800 gigabits per second from each of 144 ports, and its two CPO chips are liquid-cooled instead of air-cooled, as are an increasing fraction of new AI data centers. The network ASIC includes Nvidia’s SHARP FP8 technology, which allows CPUs and GPUs to offload certain tasks to the network chip. Spectrum-X is an Ethernet-based switch that can deliver a total bandwidth of about 100 terabits per second from a total of either 128 or 512 ports and 400 Tb/s from 512 or 2,048 ports. Hardware makers are expected to have Spectrum-X switches ready in 2026. Nvidia has been working on the fundamental photonics technology for years. But it took collaboration with 11 partners—including TSMC, Corning, and Foxconn—to get the switch to a commercial state. Ashkan Seyedi, director of optical interconnect products at Nvidia, stressed how important it was that the technologies these partners brought to the table were co-optimized to satisfy AI data-center needs rather than simply assembled from those partners’ existing technologies. “The innovations and the power savings enabled by CPO are intimately tied to your packaging scheme, your packaging partners, your packaging flow,” Seyedi says. “The novelty is not just in the optical components directly, it’s in how they are packaged in a high-yield, testable way that you can manage at good cost.” Testing is particularly important, because the system is an integration of so many expensive components. For example, there are 18 silicon photonics chiplets in each of the two CPOs in the Quantum-X system. And each of those must connect to two lasers and 16 optical fibers. Seyedi says the team had to develop several new test procedures to get it right and trace where errors were creeping in. Micas Networks Switches Micas Networks is already in production with a switch based on Broadcom’s CPO technology.Micas Network Broadcom chose the more established Mach-Zehnder modulators for its Bailly CPO switch, in part because it is a more standardized technology, potentially making it easier to integrate with existing pluggable transceiver infrastructure, explains Robert Hannah, senior manager of product marketing in Broadcom’s optical systems division. Micas’s system uses a single CPO component, which is made up of Broadcom’s Tomahawk 5 Ethernet switch chip surrounded by eight 6.4-Tb/s silicon photonics optical engines. The air-cooled hardware is in full production now, putting it ahead of Nvidia’s CPO switches. Hannah calls Nvidia’s involvement an endorsement of Micas’s and Broadcom’s timing. “Several years ago, we made the decision to skate to where the puck was going to be,” says Mitch Galbraith, Micas’s chief operations officer. With data-center operators scrambling to power their infrastructure, the CPO’s time seems to have come, he says. The new switch promises a 40 percent power savings versus systems populated with standard pluggable transceivers. However, Charlie Hou, vice president of corporate strategy at Micas, says CPO’s higher reliability is just as important. “Link flap,” the term for transient failure of pluggable optical links, is one of the culprits responsible for lengthening AI training runs that are already very long, he says. CPO is expected to have less link flap because there are fewer components in the signal’s path, among other reasons. CPOs in the Future The big power savings that data centers are looking to get from CPOs are mostly a one-time benefit, Schow suggests. After that, “I think it’s just going to be the new normal.” However, improvements to the electronics’ other features will let CPO makers keep boosting bandwidth—for a time at least. Schow doubts that individual silicon modulators—which run at 200 Gb/s in Nvidia’s photonic engines—will be able to go past much more than 400 Gb/s. However, other materials, such as lithium niobate and indium phosphide, should be able to exceed that. The trick will be affordably integrating them with silicon components, something Santa Barbara–based OpenLight is working on, among other groups. In the meantime, pluggable optics are not standing still. This week, Broadcom unveiled a new digital signal processor that could lead to a more than 20 percent power reduction for 1.6 Tb/s transceivers, due in part to a more-advanced silicon process. And startups such as Avicena, Ayar Labs, and Lightmatter are working to bring optical interconnects all the way to the GPU itself. The first two have developed chiplets meant to go inside the same package as a GPU or other processor. Lightmatter is going a step farther, making the silicon photonics engine the packaging substrate upon which future chips are 3D-stacked.

There is an inherent tension in the dissemination of research. On one hand, science thrives on openness and communication. On the other, ensuring high-quality scientific work requires peer reviews that are often lengthy and closed. In 1991, physicist Paul Ginsparg created the arXiv repository to alleviate some of that tension. The idea is that researchers have a place to upload their preprint manuscripts before they are published in a journal. The preprints are free to all but have not undergone peer review (there is some screening). However, arXiv doesn’t facilitate open, two-way discussion. Now, two Stanford students have developed an extension of arXiv that creates a centralized public square, of sorts, for researchers to discuss preprints. IEEE Spectrum spoke with one of the two, Rehaan Ahmad, about the project. ​Rehaan Ahmad Rehaan Ahmad is the cofounder of alphaXiv, which he began as an undergraduate project while at Stanford University, alongside fellow student Raj Palleti. How does alphaXiv work? Rehaan Ahmad: You can change the “arXiv” in the URL to “alphaXiv,” and it opens up the paper and there’s comments and discussion. You can highlight sections and leave in-line comments. There’s also a more general home page where you can see what papers other people are reading through the site. It ends up being a nice way to filter for what papers are interesting and what aren’t. What motivated you to create the site? Ahmad: My cocreator Raj Palleti and I were undergrads at Stanford doing research in robotics and reinforcement learning. We figured a lot of people would have questions on papers, like us. So I put together a little mock-up two or three years ago. It was just sitting on my computer for a while. And then a year afterward I showed it to Raj, and he said we need to make this a public site. We thought of it as a version of Stack Overflow for papers. How difficult was it to build? Ahmad: Surprisingly difficult! Our background is in research, and one of the harder lessons for this project is that writing research code versus actual code that works are two different things. For research code, you write something once, you put it on GitHub, no one will use it—and if they do, it’s their problem to figure out. But here, the site has been around for a year and a half, and only recently have a lot of the bugs been kind of hashed out. The project started out on a single AWS server, and anytime someone would post about it, it would go viral, and the server would go down. How do you hope alphaXiv will be used? Ahmad: I see alphaXiv as just connecting the world of research in a way that’s more productive than Twitter [now X]. People find mistakes in papers here; people will read their opinions. I have been seeing more productive discussions with the authors. Your advisors include Udacity cofounder Sebastian Thrun and Meta’s chief AI scientist, Yann LeCun. How have your advisors contributed? Ahmad: After the first few months of operating alphaXiv, we circulated a lot within the computer-science community. But after discussing the platform with [University of Maryland physics professor] Victor Galitski, we realized having his voice and opinion to guide decisions that were relevant to the physics community would be incredibly important. Those interested in computer-science papers are usually more interested in the trending/likes/filtering aspect of our site, whereas those interested in physics are usually more discussion-oriented. This article appears in the April 2025 issue as “5 Questions for Rehaan Ahmad.”

This sponsored article is brought to you by Master Bond. Master Bond EP112 is an ultra-low-viscosity, electrically insulating, two-component heat curable epoxy system designed for demanding applications requiring optical clarity and resistance to chemicals commonly used in silicon processing. This article introduces a two-part case study involving a microelectronics fabrication, showcasing EP112’s role in bonding a silicon wafer to a glass substrate. Part 1: The START Process and EP112’s Role In the first part of this case study, researchers at Lawrence Livermore National Laboratory (LLNL) developed an innovative Silicon-on-Insulator (SOI) process called START (Silicon Transfer to Arbitrary Substrate). This method enables the transformation of standard bulk silicon wafers with completed circuits into SOI-like configurations without significantly increasing manufacturing costs. By using conventional fabrication techniques, the START process combines the benefits of bulk silicon electronics with those of SOI technology while maintaining cost efficiency. A critical step in this process involved bonding a silicon wafer to a glass support substrate. EP112 was selected as the adhesive of choice due to its ultra-low viscosity, strong bonding capabilities, and high chemical resistance. The bonded structure ultimately contributed to the successful development of a prototype liquid crystal display (LCD), demonstrating EP112’s effectiveness in microelectronics fabrication. Part 2: CMOS Wafer Thinning for SEU Resistance In the second part of this study, LLNL researchers applied EP112 in a novel wafer-thinning process to enhance the reliability of CMOS-based integrated circuits (ICs). The objective was to reduce susceptibility to Single Event Upsets (SEUs) by significantly decreasing the charge collection volume within the silicon substrate. To achieve this, EP112 was used to bond two substrates together, ensuring a secure attachment throughout the wafer-thinning steps. The process involved a high-temperature alkaline etching step, where EP112’s superior chemical resistance played a crucial role in preventing de-bonding. By maintaining structural integrity under these harsh conditions, EP112 enabled the successful completion of the thinning process, further demonstrating its suitability for advanced semiconductor applications. To read more about the key parameters and requirements, and learn about the results, please download the full case study here.

Analog/RF IC design has been traditionally considered an art – sometimes even a “black art” – because, contrary to digital IC design, analog/RF design combines complexity, non-linearity, conflicting design objectives and limited automation in EDA tooling. Analog/RF IC designers rely on a blend of technical expertise, intuition, accumulated experience, and creativity to meet the demanding targets of modern applications like high operating frequencies, low power, miniaturization, and shrinking design cycles. In this webinar you will learn about the revolutionary AI-driven electromagnetic-aware methodology that automates the optimization of the floor plan of analog and RF physical layouts. Join us to discover how Ansys AI solutions can add structure to the “madness” of analog/RF design. What Attendees will Learn: How the Ansys AI-driven electromagnetic-aware methodology blends with your existing custom IC design methodology and design flow How to define goals and constraints to identify the global optimum instead of settling for locally optimal solutions How the Ansys AI-driven optimization methodology will help you shave off several days or weeks from your design cycle Register now for this free webinar!

When Meta announced its plans for a vast new fiber optic network covering 50,000 kilometers and linking five continents last month, the company’s selling point was cutting-edge undersea cable tech. What went unsaid, however, was the geopolitical challenges the project might also face, along with potential insights it could reveal about Meta’s upcoming priorities. The company is hardly alone as a private player extending long fiber optic routes across oceans. Last year Google, for instance, announced a US $1 billion investment in undersea cables connecting the United States to Japan. Titans like Meta and Google investing heavily in undersea cables represents “a trend we’ve been tracking for over a decade,” says Lane Burdette, senior analyst at the Washington, D.C.–based firm TeleGeography. The challenge comes in piecing together technical details for each project, given the inevitably sketchy notes a company’s PR team provides. (Contacted by IEEE Spectrum, a Meta spokesperson declined to comment.) Meta’s new cable will be called Waterworth, after a pioneering Meta engineer who passed away last year. Waterworth hasn’t yet been added to TeleGeography’s comprehensive global submarine cable map, Burdette says, because no geographical routing plans for the fiber network have yet been announced. Once added, it would join 81 other currently planned cable routes that TeleGeography does track across the planet, alongside the world’s other 570 undersea fiber optic cables now in service. Meta’s Next 24-Fiber-Pair Undersea Line To help contextualize Meta’s news, says Howard Kidorf, managing partner at the Hoboken, N.J.–based analysis firm Pioneer Consulting, consider a point of reference: Laying cable from California to Singapore requires some 16,000 km of fiber. But going much beyond 16,000 km, he says, pushes the limits of cable tech today. “You lose capacity on each fiber pair as you go further,” he says. “So I could say 20,000 km, but then you’re running into an economic trade-off—losing total capacity.” Tiny fiber optic amplifiers are typically built into the housings of undersea cables today. And powering that network of amplifiers can represent a real bottleneck constraining the maximum length of any given cable. “It sounds like not a very challenging thing just to put more fibers in a cable,” Kidorf says. “But it’s also a bigger challenge to be able to put more optical amplifiers in.… And the biggest challenge on top of that is how do you power those optical amplifiers?” Every 50 to 80 km, an optical amplifier inside the cable must boost the optical signal, according to Kidorf. Meanwhile, each repeater typically consumes 50 to 100 watts. Do the math, and at minimum a California-to-Singapore line needs at least 10 kilowatts coursing through it just to keep the lights on. (Real-world figures, Kidorf says, come out closer to 15 to 18 kW.) “Unrepeatered cables can have over 100 fiber pairs across a single segment,” Burdette says. “But so far, the maximum fiber pairs used in a repeatered system is 24.” Waterworth will be using all 24 fiber pairs of that present-day capacity. Which puts it at the forefront of undersea cable tech today—although Waterworth isn’t the first undersea 24-fiber cable Meta has laid down. “Meta is expected to activate Anjana, the first 24-pair repeatered system, this year,” adds Burdette. “Anjana was supplied by NEC.” (Other 24-pair fiber cables with repeaters in them are also under development both by NEC and others, Burdette notes, although Meta now appears to be first in line to actually activate such a system.) Anjana is less than 8,000 km—connecting Myrtle Beach, S.C., to Santander, Spain. It will yield the social media behemoth 480 terabits per second of new bandwidth between the United States and Europe. Compared to the hypothetical California-to-Singapore cable, above, whose 16,000-km length would stretch existing fiber-tech capabilities to the extreme, Anjana isn’t setting any underwater distance records. On the other hand, Waterworth’s anticipated 50,000-km span—more than six times that of Anjana—would represent quite a leap forward. Perhaps that is why both Kidorf and Burdette wanted to clarify something about that 50,000 figure. “50,000 is a nice headline number,” Kidorf says. “It is a lot of cable. It’s roughly the output of a single cable factory for an entire year.... But this is not one cable that goes 50,000 kilometers. It’s a cable that lands in a number of places for regeneration.” “Waterworth is one project with multiple cable systems,” Burdette says. “This distinction can get kind of muddy as cable systems often have multiple segments that may even enter service at different times. So what makes something ‘one cable’ can come down to an issue of branding.” Where Will Waterworth Make Landfall? One outstanding Waterworth question, Kidorf says, concerns where and why the undersea cable will make landfall at its six or more landing points—according to Meta’s preliminary map (above). According to Kidorf, geopolitics and tech collide where international hotspots are concerned. Nobody wants their expensive cable being damaged, either intentionally or accidentally, in a conflict zone. “For example, connectivity to get from Asia to North America without going through the Red Sea is a major goal of everybody,” Kidorf says. Another goal, he adds, concerns avoiding the South China Sea. In other words, it might be charitable to imagine Meta’s Brazilian, South African, and Indian landing points as a play to bridge the digital divide. But it’s probably not coincidence, Kidorf says, that Waterworth’s projected route also neatly circumnavigates the globe while still avoiding both of those two geopolitical tinderboxes. What doesn’t yet make sense, he adds, is how Waterworth might “unlock AI innovation” (in the words of Meta’s press release) via these particular landing points. Because AI implies big data centers awaiting the cord coming out of the ocean. Yet at least two inferred Waterworth landing points (from the approximate circles on Meta’s map) currently lack major Meta data centers, he says. “Building data centers is a more significant investment in capital than building these cables are,” Kidorf says. “So not only do you need to build a data center, you have to find a way to power them. And India is a tough place to get 500 megawatts, which is what data centers are being built out as. Brazil also is not a data center capital.” More Waterworth details will clearly be needed, that is, not only to place Waterworth on TeleGeography’s map but also to determine how the cable’s networking potential will be used—as well as how truly cutting edge Waterworth’s tech specs may actually be. “They didn’t provide enough detail to really say whether it’s a technological marvel or not, because the issue is how far can you go before you have to hit land?” Kidorf says. And returning to solid ground, he says, is the ultimate technological constraint.

Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE Spectrum robotics. We also post a weekly calendar of upcoming robotics events for the next few months. Please send us your events for inclusion. European Robotics Forum: 25–27 March 2025, STUTTGART, GERMANY RoboSoft 2025: 23–26 April 2025, LAUSANNE, SWITZERLAND ICUAS 2025: 14–17 May 2025, CHARLOTTE, N.C. ICRA 2025: 19–23 May 2025, ATLANTA London Humanoids Summit: 29–30 May 2025, LONDON IEEE RCAR 2025: 1–6 June 2025, TOYAMA, JAPAN 2025 Energy Drone & Robotics Summit: 16–18 June 2025, HOUSTON RSS 2025: 21–25 June 2025, LOS ANGELES ETH Robotics Summer School: 21–27 June 2025, GENEVA IAS 2025: 30 June–4 July 2025, GENOA, ITALY ICRES 2025: 3–4 July 2025, PORTO, PORTUGAL IEEE World Haptics: 8–11 July 2025, SUWON, KOREA IFAC Symposium on Robotics: 15–18 July 2025, PARIS RoboCup 2025: 15–21 July 2025, BAHIA, BRAZIL Enjoy today’s videos! Every time you see a humanoid demo in a warehouse or factory, ask yourself: Would a “superhumanoid” like this actually be a better answer? [ Dexterity ] The only reason that this is the second video in Video Friday this week, and not the first, is because you’ve almost certainly already seen it. This is a collaboration between the Robotics and AI Institute and Boston Dynamics, and RAI has its own video, which is slightly different: - YouTube [ Boston Dynamics ] via [ RAI ] Well this just looks a little bit like magic. [ University of Pennsylvania Sung Robotics Lab ] After hours of dance battles with professional choreographers (yes, real human dancers!), PM01 now nails every iconic move from Kung Fu Hustle. [ EngineAI ] Sanctuary AI has demonstrated industry-leading sim-to-real transfer of learned dexterous manipulation policies for our unique, high degree-of-freedom, high strength, and high speed hydraulic hands. [ Sanctuary AI ] This video is “introducing BotQ, Figure’s new high-volume manufacturing facility for humanoid robots,” but I just see some injection molding and finishing of a few plastic parts. [ Figure ] DEEP Robotics recently showcased its “One-Touch Navigation” feature, enhancing the intelligent control experience of its robotic dog. This feature offers two modes: map-based point selection and navigation and video-based point navigation, designed for open terrains and confined spaces respectively. By simply typing on a tablet screen or selecting a point in the video feed, the robotic dog can autonomously navigate to the target point, automatically planning its path and intelligently avoiding obstacles, significantly improving traversal efficiency. What’s in the bags, though? [ Deep Robotics ] This hurts my knees to watch, in a few different ways. [ Unitree ] Why the recent obsession with two legs when instead robots could have six? So much cuter! [ Jizai ] via [ RobotStart ] The world must know: who killed Mini-Duck? [ Pollen ] Seven hours of Digit robots at work at ProMat. And there are two more days of these livestreams if you need more! [ Agility ]

In partnership with Google, the Computer History Museum (CHM) has released the source code to AlexNet, the neural network that in 2012 kickstarted today’s prevailing approach to AI. The source code is available as open source on CHM’s GitHub page. What Is AlexNet? AlexNet is an artificial neural network created to recognize the contents of photographic images. It was developed in 2012 by then–University of Toronto graduate students Alex Krizhevsky and Ilya Sutskever and their faculty advisor, Geoffrey Hinton. The Origins of Deep Learning Hinton is regarded as one of the fathers of deep learning, the type of artificial intelligence that uses neural networks and is the foundation of today’s mainstream AI. Simple three-layer neural networks with only one layer of adaptive weights were first built in the late 1950s—most notably by Cornell researcher Frank Rosenblatt—but they were found to have limitations. [This explainer gives more details on how neural networks work.] In particular, researchers needed networks with more than one layer of adaptive weights, but there wasn’t a good way to train them. By the early 1970s, neural networks had been largely rejected by AI researchers. Frank Rosenblatt [left, shown with Charles W. Wightman] developed the first artificial neural network, the perceptron, in 1957.Division of Rare and Manuscript Collections/Cornell University Library In the 1980s, neural network research was revived outside the AI community by cognitive scientists at the University of California, San Diego, under the new name of “connectionism.” After finishing his Ph.D. at the University of Edinburgh in 1978, Hinton had become a postdoctoral fellow at UCSD, where he collaborated with David Rumelhart and Ronald Williams. The three rediscovered the backpropagation algorithm for training neural networks, and in 1986 they published two papers showing that it enabled neural networks to learn multiple layers of features for language and vision tasks. Backpropagation, which is foundational to deep learning today, uses the difference between the current output and the desired output of the network to adjust the weights in each layer, from the output layer backward to the input layer. In 1987, Hinton joined the University of Toronto. Away from the centers of traditional AI, Hinton’s work and those of his graduate students made Toronto a center of deep learning research over the coming decades. One postdoctoral student of Hinton’s was Yann LeCun, now chief scientist at Meta. While working in Toronto, LeCun showed that when backpropagation was used in “convolutional” neural networks, they became very good at recognizing handwritten numbers. ImageNet and GPUs Despite these advances, neural networks could not consistently outperform other types of machine learning algorithms. They needed two developments from outside of AI to pave the way. The first was the emergence of vastly larger amounts of data for training, made available through the Web. The second was enough computational power to perform this training, in the form of 3D graphics chips, known as GPUs. By 2012, the time was ripe for AlexNet. Fei-Fei Li’s ImageNet image dataset, completed in 2009, was pivotal in training AlexNet. Here, Li [right] talks with Tom Kalil at the Computer History Museum.Douglas Fairbairn/Computer History Museum The data needed to train AlexNet was found in ImageNet, a project started and led by Stanford professor Fei-Fei Li. Beginning in 2006, and against conventional wisdom, Li envisioned a dataset of images covering every noun in the English language. She and her graduate students began collecting images found on the Internet and classifying them using a taxonomy provided by WordNet, a database of words and their relationships to each other. Given the enormity of their task, Li and her collaborators ultimately crowdsourced the task of labeling images to gig workers, using Amazon’s Mechanical Turk platform. Completed in 2009, ImageNet was larger than any previous image dataset by several orders of magnitude. Li hoped its availability would spur new breakthroughs, and she started a competition in 2010 to encourage research teams to improve their image-recognition algorithms. But over the next two years, the best systems only made marginal improvements. The second condition necessary for the success of neural networks was economical access to vast amounts of computation. Neural-network training involves a lot of repeated matrix multiplications, preferably done in parallel, something that GPUs are designed to do. Nvidia, cofounded by CEO Jensen Huang, had led the way in the 2000s in making GPUs more generalizable and programmable for applications beyond 3D graphics, especially with the CUDA programming system released in 2007. Both ImageNet and CUDA were, like neural networks themselves, fairly niche developments that were waiting for the right circumstances to shine. In 2012, AlexNet brought together these elements—deep neural networks, big datasets, and GPUs—for the first time, with pathbreaking results. Each of these needed the other. How AlexNet Was Created By the late 2000s, Hinton’s grad students at the University of Toronto were beginning to use GPUs to train neural networks for both image and speech recognition. Their first successes came in speech recognition, but success in image recognition would point to deep learning as a possible general-purpose solution to AI. One student, Ilya Sutskever, believed that the performance of neural networks would scale with the amount of data available, and the arrival of ImageNet provided the opportunity. A milestone occurred in 2011, when DanNet, a convolutional neural network trained on GPUs created by Dan Cireşan and others at Jürgen Schmidhuber’s lab in Switzerland, won 4 image recognition contests. However, these results were on smaller datasets and weren’t able to move the field of computer vision. ImageNet, which was much larger and more comprehensive, was different. That same year, Sutskever convinced fellow Toronto grad student Alex Krizhevsky, who had a keen ability to wring maximum performance out of GPUs, to train a convolutional neural network for ImageNet, with Hinton serving as principal investigator. AlexNet used Nvidia GPUs running CUDA code trained on the ImageNet dataset. Nvidia CEO Jensen Huang was named a 2024 CHM Fellow for his contributions to computer graphics chips and AI.Douglas Fairbairn/Computer History Museum Krizhevsky had already written CUDA code for a convolutional neural network using Nvidia GPUs, called cuda-convnet, trained on the much smaller CIFAR-10 image dataset. He extended cuda-convnet with support for multiple GPUs and other features and retrained it on ImageNet. The training was done on a computer with two Nvidia cards in Krizhevsky’s bedroom at his parents’ house. Over the course of the next year, he constantly tweaked the network’s parameters and retrained it until it achieved performance superior to its competitors. The network would ultimately be named AlexNet, after Krizhevsky. Geoff Hinton summed up the AlexNet project this way: “Ilya thought we should do it, Alex made it work, and I got the Nobel prize.” Krizhevsky, Sutskever, and Hinton wrote a paper on AlexNet that was published in the fall of 2012 and presented by Krizhevsky at a computer-vision conference in Florence, Italy, in October. Veteran computer-vision researchers weren’t convinced, but LeCun, who was at the meeting, pronounced it a turning point for AI. He was right. Before AlexNet, almost none of the leading computer-vision papers used neural nets. After it, almost all of them would. AlexNet was just the beginning. In the next decade, neural networks would advance to synthesize believable human voices, beat champion Go players, and generate artwork, culminating with the release of ChatGPT in November 2022 by OpenAI, a company cofounded by Sutskever. Releasing the AlexNet Source Code In 2020, I reached out to Krizhevsky to ask about the possibility of allowing CHM to release the AlexNet source code, due to its historical significance. He connected me to Hinton, who was working at Google at the time. Google owned AlexNet, having acquired DNNresearch, the company owned by Hinton, Sutskever, and Krizhevsky. Hinton got the ball rolling by connecting CHM to the right team at Google. CHM worked with the Google team for five years to negotiate the release. The team also helped us identify the specific version of the AlexNet source code to release—there have been many versions of AlexNet over the years. There are other repositories of code called AlexNet on GitHub, but many of these are re-creations based on the famous paper, not the original code. CHM is proud to present the source code to the 2012 version of AlexNet, which transformed the field of artificial intelligence. You can access the source code on CHM’s GitHub page. This post originally appeared on the blog of the Computer History Museum. This article was updated on 25 March 2025. Acknowledgments Special thanks to Geoffrey Hinton for providing his quote and reviewing the text, to Cade Metz and Alex Krizhevsky for additional clarifications, and to David Bieber and the rest of the team at Google for their work in securing the source code release. References Fei-Fei Li, The Worlds I See: Curiosity, Exploration, and Discovery at the Dawn of AI. First edition, Flatiron Books, New York, 2023. Cade Metz, Genius Makers: The Mavericks Who Brought AI to Google, Facebook, and the World. First edition, Penguin Random House, New York, 2022.

Technology should benefit humanity. One of the most remarkable examples of technology’s potential to provide enduring benefits is the Itaipu Hydroelectric Dam, a massive binational energy project between Brazil and Paraguay. Built on the Paraná River, which forms part of the border between the two nations, Itaipu transformed a once-contested hydroelectric resource into a shared engine of economic progress. The power plant has held many records. For decades, it was the world’s largest hydroelectric facility; the dam spans the river’s 7.9-kilometer width and reaches a height of 196 meters. Itaipu was also the first hydropower plant to generate more than 100 terawatt hours of electricity in a year. To acknowledge Itaipu’s monumental engineering achievement, on 27 March the dam will be recognized as an IEEE Milestone during a ceremony in Hernandarias, Paraguay. The ceremony will commemorate the project’s impact on engineering and energy production. Itaipu’s massive scale By the late 1960s, Brazil and Paraguay recognized the Paraná River’s untapped hydroelectric potential, according to the Global Infrastructure Hub. Brazil, which was undergoing rapid industrialization, sought a stable, renewable energy source to reduce its dependence on fossil fuels. Meanwhile, Paraguay, lacking the financial resources to construct a gigawatt-scale hydroelectric facility independently, entered into a treaty with Brazil in 1973. The agreement granted both countries equal ownership of the dam and its power generation. Construction began in 1975 and was completed in 1984, costing US $19.6 billion. The scale of the project was staggering. Engineers excavated 50 million cubic meters of earth and rock, poured 12.3 million cubic meters of concrete, and used enough iron and steel to construct 380 Eiffel Towers. Itaipu was designed for continuous expansion. It initially launched with two 700-megawatt turbine units, providing 1.4 gigawatts of capacity. By 1991, the power plant reached its planned 12.6 GW capacity. In 2006 and 2007, it was expanded to 14 GW with the addition of two more units, for a total of 20. Although China’s 22.5-GW Three Gorges Dam, on the Yangtze River near the city of Yichang, surpassed Itaipu’s capacity in 2012, the South American dam remains one of the world’s most productive hydroelectric facilities. On average, Itaipu generates around 90 terawatt-hours of electricity annually. It set a record by generating 103.1 TWh in 2016 (surpassed in 2020 by Three Gorges’ 111.8-TWh output). To put 100 TWh into perspective, a power plant would need to burn approximately 50 million tonnes of coal to produce the same amount of energy, according to the U.S. Energy Information Administration. By harnessing 62,200 cubic meters of river water per second, Itaipu prevents the release of nearly 100 million tonnes of carbon dioxide each year. During its 40-year lifetime, the dam has generated more than 3,000 TWh of electricity, meeting nearly 90 percent of Paraguay’s energy needs and contributing roughly 10 percent of Brazil’s electricity supply. Itaipu’s legacy endures as a testament to the benefits of international cooperation and sustainable energy and to the power of engineering to shape the future. IEEE recognition for Itaipu The IEEE Milestone commemorative plaque, now displayed in the dam’s visitor center, highlights Itaipu’s role as a world leader in hydroelectric power generation. It reads: “Itaipu power plant construction began in 1975 as a joint Brazil-Paraguay venture. When power generation started in 1984, Itaipu set a world record for the single largest installed hydroelectric capacity (14 GW). For at least three decades, Itaipu produced more electricity annually than any other hydroelectric project. Linking power plants, substations, and transmission lines in both Brazil and Paraguay, Itaipu’s system provided reliable, affordable energy to consumers and industry.” Administered by the IEEE History Center and supported by donors, the Milestone program recognizes outstanding technical developments worldwide. The IEEE Paraguay Section sponsored the nomination. This article was updated on 21 March 2025.

The rapid development of AI is fueling a data center boom, unlocking billions of dollars in investments to build the infrastructure needed to support data- and energy-hungry models. Amazon, Microsoft, and Google are among the key players backing large-scale AI projects, betting that new data centers will create jobs. In the United States, the Trump administration announced in late January the US $500 billion Stargate Project, a partnership with OpenAI, Oracle, and SoftBank to build data centers nationwide. The project promises more than 100,000 new U.S. jobs over the next four years. Globally, the demand for data centers is projected to rise between 19 and 22 percent from 2023 to 2030, according to McKinsey & Co., with operators expanding their facilities into parts of Asia, South America, and the Middle East. This surge in construction has created a strong demand for certain electrical engineers, whose expertise in power systems and energy efficiency is essential for designing, building, and maintaining energy-intensive AI infrastructure. The data center industry contributed 4.7 million jobs to the U.S. economy in 2023—a 60 percent increase from 2017, according to a 2025 PwC report. The U.S. Bureau of Labor Statistics projects a 9 percent growth rate for electrical engineering jobs between 2023 and 2033, more than double the average for all occupations, with a median pay of $109,010 per year. But with the uptick in demand, data centers are struggling to find qualified talent. A 2023 Uptime Institute report found that 58 percent of global data center operators faced difficulties sourcing talent for open roles. Without enough skilled engineers to manage power-distribution systems, data centers risk increased downtime and potential grid instability, especially as AI-driven workloads demand more energy. Engineers Are Key to AI Infrastructure Big tech companies are increasingly searching for electrical engineers to scale their data infrastructure. As of mid-March, Amazon, Meta, and Google were looking to fill roles for electrical design engineers, research and development engineers, and mechatronics engineers to build robotic systems, with salaries reaching $281,000 a year, according to job listings. Oracle, which operates more than 150 data centers worldwide, has listed more than a dozen electrical engineering roles across its data facilities. Engineering firms are also seeing a spike in demand. Karina Hershberg, associate principal at PAE Consulting Engineers, says data center clients in need of her firm’s expertise have drastically increased over the past two years. Many clients require power systems on the scale of a “small town,” Hershberg says, and electrical engineers are expected to meet their demands within tight timelines of a year or two. The work involves sourcing power, optimizing energy distribution, designing emergency backup systems, and integrating cooling mechanisms to prevent overheating. As AI advances, engineers must also address new challenges in power stability and energy efficiency. “It’s all hands on deck,” Hershberg says. “AI is introducing a whole new set of challenges to power systems, and we need people who really understand the engineering and science behind it.” Sustainability efforts are further boosting demand for electrical engineers. As data centers explore solar, wind, and nuclear energy, engineers must design power-distribution systems that can efficiently transmit energy from production sites to facilities, according to Jim Kozlowski, chief sustainability officer and vice president of global data center operations at Ensono, an IT service management firm. Engineers specializing in renewable energy, he says, will find plenty of opportunities in the data center industry. “As an engineer coming into this world, you could focus on renewable energy and develop a great career there, because those opportunities are only going to grow,” Kozlowski says. Competitive Data Center Jobs The hiring landscape for electrical engineers in data centers is fiercely competitive. Grace Søhoel-Goldberg, who leads electrical engineering recruitment at LVI Associates, a global staffing firm for the infrastructure sector, has seen demand for electrical engineers soar as more companies enter the data center ring. What was once a small group of companies competing for niche talent, she says, has nearly tripled over the past few years. “It’s very cutthroat,” Søhoel-Goldberg says of the data center job market. Even with high demand, hiring the right people remains a challenge, she adds. Data center operators and engineering firms compete for the same limited pool of candidates, and large companies with better salaries and perks have the advantage, often poaching talent from smaller firms. Another barrier is employer expectations. Many companies only want to hire engineers who’ve been working at data centers for years, making it harder to hire qualified candidates from transferable backgrounds such as industrial engineering. Søhoel-Goldberg believes this creates an unnecessary talent bottleneck. A lack of awareness about data center careers further widens the talent gap. Electrical engineering roles in data centers require specialized knowledge in power infrastructure, typically learned in the construction industry. But based on conversations with students, PAE Consulting’s Hershberg says that university programs rarely highlight career paths in this sector. As a result, graduates are often steered toward software and high-tech fields, making it harder for them to transition into data centers later in their careers. “We’re just not seeing the talent pool at that graduation level,” Hershberg says. Closing the Talent Gap Some universities are stepping up to address the skills shortage. Southern Methodist University in Texas, for example, offers one of the few master’s programs in data center systems engineering designed for students interested in a data center career. The University of Wisconsin–Madison runs boot camps including a three-day course on data center design and operation, preparing students for roles such as electrical designers. Beyond universities, online platforms like Udemy offer data center fundamentals courses, and Amazon Web Services and Microsoft run skills training programs for both current employees and aspiring professionals to create a pipeline of entry-level technician talent. From a hiring perspective, Søhoel-Goldberg advises engineers without direct data center experience to gain transferable skills in uninterruptible power supply (UPS) systems by working in other sectors including health care, wastewater treatment, or manufacturing. She also recommends acquiring certifications such as Engineer-in-Training (EIT) and Professional Engineer (PE) licenses.

Andrew Ng has serious street cred in artificial intelligence. He pioneered the use of graphics processing units (GPUs) to train deep learning models in the late 2000s with his students at Stanford University, cofounded Google Brain in 2011, and then served for three years as chief scientist for Baidu, where he helped build the Chinese tech giant’s AI group. So when he says he has identified the next big shift in artificial intelligence, people listen. And that’s what he told IEEE Spectrum in an exclusive Q&A. Ng’s current efforts are focused on his company Landing AI, which built a platform called LandingLens to help manufacturers improve visual inspection with computer vision. He has also become something of an evangelist for what he calls the data-centric AI movement, which he says can yield “small data” solutions to big issues in AI, including model efficiency, accuracy, and bias. Andrew Ng on... What’s next for really big models The career advice he didn’t listen to Defining the data-centric AI movement Synthetic data Why Landing AI asks its customers to do the work The great advances in deep learning over the past decade or so have been powered by ever-bigger models crunching ever-bigger amounts of data. Some people argue that that’s an unsustainable trajectory. Do you agree that it can’t go on that way? Andrew Ng: This is a big question. We’ve seen foundation models in NLP [natural language processing]. I’m excited about NLP models getting even bigger, and also about the potential of building foundation models in computer vision. I think there’s lots of signal to still be exploited in video: We have not been able to build foundation models yet for video because of compute bandwidth and the cost of processing video, as opposed to tokenized text. So I think that this engine of scaling up deep learning algorithms, which has been running for something like 15 years now, still has steam in it. Having said that, it only applies to certain problems, and there’s a set of other problems that need small data solutions. When you say you want a foundation model for computer vision, what do you mean by that? Ng: This is a term coined by Percy Liang and some of my friends at Stanford to refer to very large models, trained on very large data sets, that can be tuned for specific applications. For example, GPT-3 is an example of a foundation model [for NLP]. Foundation models offer a lot of promise as a new paradigm in developing machine learning applications, but also challenges in terms of making sure that they’re reasonably fair and free from bias, especially if many of us will be building on top of them. What needs to happen for someone to build a foundation model for video? Ng: I think there is a scalability problem. The compute power needed to process the large volume of images for video is significant, and I think that’s why foundation models have arisen first in NLP. Many researchers are working on this, and I think we’re seeing early signs of such models being developed in computer vision. But I’m confident that if a semiconductor maker gave us 10 times more processor power, we could easily find 10 times more video to build such models for vision. Having said that, a lot of what’s happened over the past decade is that deep learning has happened in consumer-facing companies that have large user bases, sometimes billions of users, and therefore very large data sets. While that paradigm of machine learning has driven a lot of economic value in consumer software, I find that that recipe of scale doesn’t work for other industries. Back to top It’s funny to hear you say that, because your early work was at a consumer-facing company with millions of users. Ng: Over a decade ago, when I proposed starting the Google Brain project to use Google’s compute infrastructure to build very large neural networks, it was a controversial step. One very senior person pulled me aside and warned me that starting Google Brain would be bad for my career. I think he felt that the action couldn’t just be in scaling up, and that I should instead focus on architecture innovation. “In many industries where giant data sets simply don’t exist, I think the focus has to shift from big data to good data. Having 50 thoughtfully engineered examples can be sufficient to explain to the neural network what you want it to learn.” —Andrew Ng, CEO & Founder, Landing AI I remember when my students and I published the first NeurIPS workshop paper advocating using CUDA, a platform for processing on GPUs, for deep learning—a different senior person in AI sat me down and said, “CUDA is really complicated to program. As a programming paradigm, this seems like too much work.” I did manage to convince him; the other person I did not convince. I expect they’re both convinced now. Ng: I think so, yes. Over the past year as I’ve been speaking to people about the data-centric AI movement, I’ve been getting flashbacks to when I was speaking to people about deep learning and scalability 10 or 15 years ago. In the past year, I’ve been getting the same mix of “there’s nothing new here” and “this seems like the wrong direction.” Back to top How do you define data-centric AI, and why do you consider it a movement? Ng: Data-centric AI is the discipline of systematically engineering the data needed to successfully build an AI system. For an AI system, you have to implement some algorithm, say a neural network, in code and then train it on your data set. The dominant paradigm over the last decade was to download the data set while you focus on improving the code. Thanks to that paradigm, over the last decade deep learning networks have improved significantly, to the point where for a lot of applications the code—the neural network architecture—is basically a solved problem. So for many practical applications, it’s now more productive to hold the neural network architecture fixed, and instead find ways to improve the data. When I started speaking about this, there were many practitioners who, completely appropriately, raised their hands and said, “Yes, we’ve been doing this for 20 years.” This is the time to take the things that some individuals have been doing intuitively and make it a systematic engineering discipline. The data-centric AI movement is much bigger than one company or group of researchers. My collaborators and I organized a data-centric AI workshop at NeurIPS, and I was really delighted at the number of authors and presenters that showed up. You often talk about companies or institutions that have only a small amount of data to work with. How can data-centric AI help them? Ng: You hear a lot about vision systems built with millions of images—I once built a face recognition system using 350 million images. Architectures built for hundreds of millions of images don’t work with only 50 images. But it turns out, if you have 50 really good examples, you can build something valuable, like a defect-inspection system. In many industries where giant data sets simply don’t exist, I think the focus has to shift from big data to good data. Having 50 thoughtfully engineered examples can be sufficient to explain to the neural network what you want it to learn. When you talk about training a model with just 50 images, does that really mean you’re taking an existing model that was trained on a very large data set and fine-tuning it? Or do you mean a brand new model that’s designed to learn only from that small data set? Ng: Let me describe what Landing AI does. When doing visual inspection for manufacturers, we often use our own flavor of RetinaNet. It is a pretrained model. Having said that, the pretraining is a small piece of the puzzle. What’s a bigger piece of the puzzle is providing tools that enable the manufacturer to pick the right set of images [to use for fine-tuning] and label them in a consistent way. There’s a very practical problem we’ve seen spanning vision, NLP, and speech, where even human annotators don’t agree on the appropriate label. For big data applications, the common response has been: If the data is noisy, let’s just get a lot of data and the algorithm will average over it. But if you can develop tools that flag where the data’s inconsistent and give you a very targeted way to improve the consistency of the data, that turns out to be a more efficient way to get a high-performing system. “Collecting more data often helps, but if you try to collect more data for everything, that can be a very expensive activity.” —Andrew Ng For example, if you have 10,000 images where 30 images are of one class, and those 30 images are labeled inconsistently, one of the things we do is build tools to draw your attention to the subset of data that’s inconsistent. So you can very quickly relabel those images to be more consistent, and this leads to improvement in performance. Could this focus on high-quality data help with bias in data sets? If you’re able to curate the data more before training? Ng: Very much so. Many researchers have pointed out that biased data is one factor among many leading to biased systems. There have been many thoughtful efforts to engineer the data. At the NeurIPS workshop, Olga Russakovsky gave a really nice talk on this. At the main NeurIPS conference, I also really enjoyed Mary Gray’s presentation, which touched on how data-centric AI is one piece of the solution, but not the entire solution. New tools like Datasheets for Datasets also seem like an important piece of the puzzle. One of the powerful tools that data-centric AI gives us is the ability to engineer a subset of the data. Imagine training a machine-learning system and finding that its performance is okay for most of the data set, but its performance is biased for just a subset of the data. If you try to change the whole neural network architecture to improve the performance on just that subset, it’s quite difficult. But if you can engineer a subset of the data you can address the problem in a much more targeted way. When you talk about engineering the data, what do you mean exactly? Ng: In AI, data cleaning is important, but the way the data has been cleaned has often been in very manual ways. In computer vision, someone may visualize images through a Jupyter notebook and maybe spot the problem, and maybe fix it. But I’m excited about tools that allow you to have a very large data set, tools that draw your attention quickly and efficiently to the subset of data where, say, the labels are noisy. Or to quickly bring your attention to the one class among 100 classes where it would benefit you to collect more data. Collecting more data often helps, but if you try to collect more data for everything, that can be a very expensive activity. For example, I once figured out that a speech-recognition system was performing poorly when there was car noise in the background. Knowing that allowed me to collect more data with car noise in the background, rather than trying to collect more data for everything, which would have been expensive and slow. Back to top What about using synthetic data, is that often a good solution? Ng: I think synthetic data is an important tool in the tool chest of data-centric AI. At the NeurIPS workshop, Anima Anandkumar gave a great talk that touched on synthetic data. I think there are important uses of synthetic data that go beyond just being a preprocessing step for increasing the data set for a learning algorithm. I’d love to see more tools to let developers use synthetic data generation as part of the closed loop of iterative machine learning development. Do you mean that synthetic data would allow you to try the model on more data sets? Ng: Not really. Here’s an example. Let’s say you’re trying to detect defects in a smartphone casing. There are many different types of defects on smartphones. It could be a scratch, a dent, pit marks, discoloration of the material, other types of blemishes. If you train the model and then find through error analysis that it’s doing well overall but it’s performing poorly on pit marks, then synthetic data generation allows you to address the problem in a more targeted way. You could generate more data just for the pit-mark category. “In the consumer software Internet, we could train a handful of machine-learning models to serve a billion users. In manufacturing, you might have 10,000 manufacturers building 10,000 custom AI models.” —Andrew Ng Synthetic data generation is a very powerful tool, but there are many simpler tools that I will often try first. Such as data augmentation, improving labeling consistency, or just asking a factory to collect more data. Back to top To make these issues more concrete, can you walk me through an example? When a company approaches Landing AI and says it has a problem with visual inspection, how do you onboard them and work toward deployment? Ng: When a customer approaches us we usually have a conversation about their inspection problem and look at a few images to verify that the problem is feasible with computer vision. Assuming it is, we ask them to upload the data to the LandingLens platform. We often advise them on the methodology of data-centric AI and help them label the data. One of the foci of Landing AI is to empower manufacturing companies to do the machine learning work themselves. A lot of our work is making sure the software is fast and easy to use. Through the iterative process of machine learning development, we advise customers on things like how to train models on the platform, when and how to improve the labeling of data so the performance of the model improves. Our training and software supports them all the way through deploying the trained model to an edge device in the factory. How do you deal with changing needs? If products change or lighting conditions change in the factory, can the model keep up? Ng: It varies by manufacturer. There is data drift in many contexts. But there are some manufacturers that have been running the same manufacturing line for 20 years now with few changes, so they don’t expect changes in the next five years. Those stable environments make things easier. For other manufacturers, we provide tools to flag when there’s a significant data-drift issue. I find it really important to empower manufacturing customers to correct data, retrain, and update the model. Because if something changes and it’s 3 a.m. in the United States, I want them to be able to adapt their learning algorithm right away to maintain operations. In the consumer software Internet, we could train a handful of machine-learning models to serve a billion users. In manufacturing, you might have 10,000 manufacturers building 10,000 custom AI models. The challenge is, how do you do that without Landing AI having to hire 10,000 machine learning specialists? So you’re saying that to make it scale, you have to empower customers to do a lot of the training and other work. Ng: Yes, exactly! This is an industry-wide problem in AI, not just in manufacturing. Look at health care. Every hospital has its own slightly different format for electronic health records. How can every hospital train its own custom AI model? Expecting every hospital’s IT personnel to invent new neural-network architectures is unrealistic. The only way out of this dilemma is to build tools that empower the customers to build their own models by giving them tools to engineer the data and express their domain knowledge. That’s what Landing AI is executing in computer vision, and the field of AI needs other teams to execute this in other domains. Is there anything else you think it’s important for people to understand about the work you’re doing or the data-centric AI movement? Ng: In the last decade, the biggest shift in AI was a shift to deep learning. I think it’s quite possible that in this decade the biggest shift will be to data-centric AI. With the maturity of today’s neural network architectures, I think for a lot of the practical applications the bottleneck will be whether we can efficiently get the data we need to develop systems that work well. The data-centric AI movement has tremendous energy and momentum across the whole community. I hope more researchers and developers will jump in and work on it. Back to top This article appears in the April 2022 print issue as “Andrew Ng, AI Minimalist.”

The end of Moore’s Law is looming. Engineers and designers can do only so much to miniaturize transistors and pack as many of them as possible into chips. So they’re turning to other approaches to chip design, incorporating technologies like AI into the process. Samsung, for instance, is adding AI to its memory chips to enable processing in memory, thereby saving energy and speeding up machine learning. Speaking of speed, Google’s TPU V4 AI chip has doubled its processing power compared with that of its previous version. But AI holds still more promise and potential for the semiconductor industry. To better understand how AI is set to revolutionize chip design, we spoke with Heather Gorr, senior product manager for MathWorks’ MATLAB platform. How is AI currently being used to design the next generation of chips? Heather Gorr: AI is such an important technology because it’s involved in most parts of the cycle, including the design and manufacturing process. There’s a lot of important applications here, even in the general process engineering where we want to optimize things. I think defect detection is a big one at all phases of the process, especially in manufacturing. But even thinking ahead in the design process, [AI now plays a significant role] when you’re designing the light and the sensors and all the different components. There’s a lot of anomaly detection and fault mitigation that you really want to consider. Heather GorrMathWorks Then, thinking about the logistical modeling that you see in any industry, there is always planned downtime that you want to mitigate; but you also end up having unplanned downtime. So, looking back at that historical data of when you’ve had those moments where maybe it took a bit longer than expected to manufacture something, you can take a look at all of that data and use AI to try to identify the proximate cause or to see something that might jump out even in the processing and design phases. We think of AI oftentimes as a predictive tool, or as a robot doing something, but a lot of times you get a lot of insight from the data through AI. What are the benefits of using AI for chip design? Gorr: Historically, we’ve seen a lot of physics-based modeling, which is a very intensive process. We want to do a reduced order model, where instead of solving such a computationally expensive and extensive model, we can do something a little cheaper. You could create a surrogate model, so to speak, of that physics-based model, use the data, and then do your parameter sweeps, your optimizations, your Monte Carlo simulations using the surrogate model. That takes a lot less time computationally than solving the physics-based equations directly. So, we’re seeing that benefit in many ways, including the efficiency and economy that are the results of iterating quickly on the experiments and the simulations that will really help in the design. So it’s like having a digital twin in a sense? Gorr: Exactly. That’s pretty much what people are doing, where you have the physical system model and the experimental data. Then, in conjunction, you have this other model that you could tweak and tune and try different parameters and experiments that let sweep through all of those different situations and come up with a better design in the end. So, it’s going to be more efficient and, as you said, cheaper? Gorr: Yeah, definitely. Especially in the experimentation and design phases, where you’re trying different things. That’s obviously going to yield dramatic cost savings if you’re actually manufacturing and producing [the chips]. You want to simulate, test, experiment as much as possible without making something using the actual process engineering. We’ve talked about the benefits. How about the drawbacks? Gorr: The [AI-based experimental models] tend to not be as accurate as physics-based models. Of course, that’s why you do many simulations and parameter sweeps. But that’s also the benefit of having that digital twin, where you can keep that in mind—it’s not going to be as accurate as that precise model that we’ve developed over the years. Both chip design and manufacturing are system intensive; you have to consider every little part. And that can be really challenging. It’s a case where you might have models to predict something and different parts of it, but you still need to bring it all together. One of the other things to think about too is that you need the data to build the models. You have to incorporate data from all sorts of different sensors and different sorts of teams, and so that heightens the challenge. How can engineers use AI to better prepare and extract insights from hardware or sensor data? Gorr: We always think about using AI to predict something or do some robot task, but you can use AI to come up with patterns and pick out things you might not have noticed before on your own. People will use AI when they have high-frequency data coming from many different sensors, and a lot of times it’s useful to explore the frequency domain and things like data synchronization or resampling. Those can be really challenging if you’re not sure where to start. One of the things I would say is, use the tools that are available. There’s a vast community of people working on these things, and you can find lots of examples [of applications and techniques] on GitHub or MATLAB Central, where people have shared nice examples, even little apps they’ve created. I think many of us are buried in data and just not sure what to do with it, so definitely take advantage of what’s already out there in the community. You can explore and see what makes sense to you, and bring in that balance of domain knowledge and the insight you get from the tools and AI. What should engineers and designers consider when using AI for chip design? Gorr: Think through what problems you’re trying to solve or what insights you might hope to find, and try to be clear about that. Consider all of the different components, and document and test each of those different parts. Consider all of the people involved, and explain and hand off in a way that is sensible for the whole team. How do you think AI will affect chip designers’ jobs? Gorr: It’s going to free up a lot of human capital for more advanced tasks. We can use AI to reduce waste, to optimize the materials, to optimize the design, but then you still have that human involved whenever it comes to decision-making. I think it’s a great example of people and technology working hand in hand. It’s also an industry where all people involved—even on the manufacturing floor—need to have some level of understanding of what’s happening, so this is a great industry for advancing AI because of how we test things and how we think about them before we put them on the chip. How do you envision the future of AI and chip design? Gorr: It’s very much dependent on that human element—involving people in the process and having that interpretable model. We can do many things with the mathematical minutiae of modeling, but it comes down to how people are using it, how everybody in the process is understanding and applying it. Communication and involvement of people of all skill levels in the process are going to be really important. We’re going to see less of those superprecise predictions and more transparency of information, sharing, and that digital twin—not only using AI but also using our human knowledge and all of the work that many people have done over the years.

Quantum computing is a devilishly complex technology, with many technical hurdles impacting its development. Of these challenges two critical issues stand out: miniaturization and qubit quality. IBM has adopted the superconducting qubit road map of reaching a 1,121-qubit processor by 2023, leading to the expectation that 1,000 qubits with today’s qubit form factor is feasible. However, current approaches will require very large chips (50 millimeters on a side, or larger) at the scale of small wafers, or the use of chiplets on multichip modules. While this approach will work, the aim is to attain a better path toward scalability. Now researchers at MIT have been able to both reduce the size of the qubits and done so in a way that reduces the interference that occurs between neighboring qubits. The MIT researchers have increased the number of superconducting qubits that can be added onto a device by a factor of 100. “We are addressing both qubit miniaturization and quality,” said William Oliver, the director for the Center for Quantum Engineering at MIT. “Unlike conventional transistor scaling, where only the number really matters, for qubits, large numbers are not sufficient, they must also be high-performance. Sacrificing performance for qubit number is not a useful trade in quantum computing. They must go hand in hand.” The key to this big increase in qubit density and reduction of interference comes down to the use of two-dimensional materials, in particular the 2D insulator hexagonal boron nitride (hBN). The MIT researchers demonstrated that a few atomic monolayers of hBN can be stacked to form the insulator in the capacitors of a superconducting qubit. Just like other capacitors, the capacitors in these superconducting circuits take the form of a sandwich in which an insulator material is sandwiched between two metal plates. The big difference for these capacitors is that the superconducting circuits can operate only at extremely low temperatures—less than 0.02 degrees above absolute zero (-273.15 °C). Superconducting qubits are measured at temperatures as low as 20 millikelvin in a dilution refrigerator.Nathan Fiske/MIT In that environment, insulating materials that are available for the job, such as PE-CVD silicon oxide or silicon nitride, have quite a few defects that are too lossy for quantum computing applications. To get around these material shortcomings, most superconducting circuits use what are called coplanar capacitors. In these capacitors, the plates are positioned laterally to one another, rather than on top of one another. As a result, the intrinsic silicon substrate below the plates and to a smaller degree the vacuum above the plates serve as the capacitor dielectric. Intrinsic silicon is chemically pure and therefore has few defects, and the large size dilutes the electric field at the plate interfaces, all of which leads to a low-loss capacitor. The lateral size of each plate in this open-face design ends up being quite large (typically 100 by 100 micrometers) in order to achieve the required capacitance. In an effort to move away from the large lateral configuration, the MIT researchers embarked on a search for an insulator that has very few defects and is compatible with superconducting capacitor plates. “We chose to study hBN because it is the most widely used insulator in 2D material research due to its cleanliness and chemical inertness,” said colead author Joel Wang, a research scientist in the Engineering Quantum Systems group of the MIT Research Laboratory for Electronics. On either side of the hBN, the MIT researchers used the 2D superconducting material, niobium diselenide. One of the trickiest aspects of fabricating the capacitors was working with the niobium diselenide, which oxidizes in seconds when exposed to air, according to Wang. This necessitates that the assembly of the capacitor occur in a glove box filled with argon gas. While this would seemingly complicate the scaling up of the production of these capacitors, Wang doesn’t regard this as a limiting factor. “What determines the quality factor of the capacitor are the two interfaces between the two materials,” said Wang. “Once the sandwich is made, the two interfaces are “sealed” and we don’t see any noticeable degradation over time when exposed to the atmosphere.” This lack of degradation is because around 90 percent of the electric field is contained within the sandwich structure, so the oxidation of the outer surface of the niobium diselenide does not play a significant role anymore. This ultimately makes the capacitor footprint much smaller, and it accounts for the reduction in cross talk between the neighboring qubits. “The main challenge for scaling up the fabrication will be the wafer-scale growth of hBN and 2D superconductors like [niobium diselenide], and how one can do wafer-scale stacking of these films,” added Wang. Wang believes that this research has shown 2D hBN to be a good insulator candidate for superconducting qubits. He says that the groundwork the MIT team has done will serve as a road map for using other hybrid 2D materials to build superconducting circuits.