2026-06-19 07:35:56
Scientists are exploring new algorithms, hardware, and computing methods to lower AI’s power demands. Strategic siting of data centers and other steps to increase green energy use are also key.
This story was originally published by Knowable Magazine.
As I sip coffee in my Berlin apartment and fire a question at Google’s AI chatbot Gemini, it’s easy not to think about the energy it takes to generate a response. Once the signal reaches my router, it whizzes, I assume, through copper wires or fiber-optic cables to one of Google’s data center hubs. Somewhere inside the data center’s labyrinthine halls of stacked processors, my query gets converted into numbers and undergoes billions of computations to determine context and meaning. The answer, once assembled, races back, in the blink of an eye.
Data centers—the beating hearts of the internet, powering everything from email to web searches—have existed for decades, but with the growing popularity of AI to generate text, images, and video, they’re using more energy than ever. According to Google’s own estimates, processing a median-length text prompt with its AI assistant Gemini consumes around 0.24 watt-hours.
These amounts, individually small—0.24 watt-hours is equivalent to watching TV for about nine seconds—are adding up fast. In March 2026, OpenAI estimated that more than 900 million people use its AI chatbot, ChatGPT, every week, tallying billions of queries daily.
The exact amount of electricity consumed by data centers, globally or in the United States, which hosts more than any other nation, isn’t publicly reported by all tech companies, says Eric Masanet of the University of California, Santa Barbara, who researches data center sustainability. But according to the most recent estimates by the International Energy Agency, US data centers guzzled some 224 terawatt-hours of electricity in 2025—more than 5 percent of the country’s electricity use. That’s a significant uptick from an estimated 1.9 percent consumed in 2018, well before the mainstream surge of generative AI.
This electricity use seems set to soar. In the race to secure market leadership for generative AI products, companies like Google, Meta, Amazon, OpenAI, Anthropic, Microsoft, and Oracle are investing tens to hundreds of billions of dollars to build AI-focused data centers. Compared to data centers of the pre-AI days that consume, say, 100 megawatts of electricity—enough to power 83,000 homes with average demand—the newcomers are often “hyperscale” and can use a gigawatt or more, or roughly a tenth of the electrical capacity of Los Angeles.
Masanet and other experts have been alarmed to see much of this demand met by plants powered by fossil fuels, such as gas, whose burning releases planet-warming carbon dioxide. A key reason is that data centers are often constructed in places without abundant renewable energy sources like hydropower, geothermal, solar, or wind.
Tech companies often offset emissions by investing in renewable energy elsewhere. But unless those clean energy plants make more energy than the data centers use, this strategy—at best—keeps CO2 emissions of centers in stasis rather than reducing them to a net of nothing, important for halting global warming. “For every megawatt for which we install fossil fuel power,” Masanet says, “it sets us back on our progress.”
And that’s not considering the resources spent on manufacturing the hardware that fills new data centers, or the impacts on communities living near them, which often suffer from air and noise pollution from gas plants and possible strain on local water resources, which are used to cool the data centers.
Although forecasts for AI’s energy impact remain devilishly tricky, especially since the size of payoffs from investments in AI are uncertain, it’s clear to experts that energy-saving strategies are urgently needed. Without them, according to one 2025 estimate, US data centers could soon be releasing the equivalent of 24 to 44 megatons of CO2 annually, the latter equivalent to the annual emissions of Norway.
And so computer scientists and engineers are rethinking some of the power-hungry hardware and software that fuel AI. They’re working to develop energy-saving algorithms and processor designs, and carefully considering where, and how, data centers are constructed.
“AI’s energy cost is not an accident: This is basically a product of how our systems are built,” says Fengqi You, an expert in energy systems at Cornell University. But with the right mix of solutions, he says, “we could really reshape the trajectory.”
To comprehend AI’s energy cost, it helps to understand large language models (LLMs)—the lifeblood of AI text generation tools such as chatbots and AI assistants—specifically, ones based on a design described in 2017 by the machine-learning laboratory Google Brain. This design, transformer architecture, can process text at lightning speed by simultaneously taking each word and weighing its relationship to every other word it sees. It “learns” which words go together by computing how strongly each word relates to all other words in a text, examining each word in many contexts. (A similar design is used for AI image and video generators.)
On a computational level, this happens by converting words or word fragments into numbers and performing additions and multiplications between them. Key to the speed is being able to do these calculations in parallel, made possible by graphic processor units (GPUs)—mostly manufactured by the company Nvidia—originally invented for rapid 3D rendering of imagery during gaming.
The initial training of an LLM, required to learn all these relationships, consumes vast amounts of energy. Because each word it trains on must be weighed against all others in a given chunk of text, the number of computations the model performs—hence the energy required—increases quadratically relative to the length of text (i.e., doubling the length of text quadruples the number of computations). That adds up quickly given that most LLMs are trained on massive swaths of publicly available internet text. Some estimates suggest that training GPT-4—the iteration of ChatGPT that launched in 2023—guzzled between 50 and 60 gigawatt-hours of electricity, enough to power San Francisco for three to four days.
But experts are more worried about the energy costs of using the models to generate data once they’ve been trained, a process called inference. “You train once, then you inference for a billion people in the world,” says Mosharaf Chowdhury, an AI systems expert at the University of Michigan who has been measuring the electricity usage of a handful of large language models that have been made publicly available.
This process is surprisingly inefficient: Each time transformer models generate a word—by selecting the one with the highest probability of following the previous word, given context—they put the query and partially written answer through the model. In doing so, they apply all of the parameters they’ve calculated during training to understand language patterns—which number in the hundreds of billions or even trillions.
“The fact that you have to do a lot of calculations for a single word to be added—that’s a problematic thing,” says Günter Klambauer, an AI expert at Johannes Kepler University in Austria.
This recognition has triggered interest in smaller language models specialized to specific tasks. These are trained more narrowly, have fewer parameters—say, tens or hundreds of millions—and perform substantially less computation than larger models. In one 2025 paper published by UNESCO, computer scientist Ivana Drobnjak of University College London and colleagues compared energy consumption of Meta’s language model Llama-3.1 with smaller AI models dedicated to particular tasks—ones called DistilBART and t5-small-xsum for summarization, and others for translation or answering questions. When used for their respective tasks, the smaller models consumed more than 90 percent less energy than Llama 3.1 on the same job.
And so computer scientists have been driven to build a similar kind of task specialization into LLMs themselves. In “mixture of expert” models, only particular parts of one big model are activated for certain tasks. These parts “learn to handle different patterns in language,” Drobnjak says.
This is thought to be one reason why R1, an LLM developed by the Chinese company DeepSeek, reportedly consumed significantly less energy than other models (independent experts have raised doubts about those figures). Udit Gupta, an expert in electrical and computer engineering at Cornell Tech, says that LLMs like Gemini or ChatGPT are similarly routing queries to more specialized sub-models. “There’s a lot of work being done on how to assess the complexity of the query or task that’s coming from users and then find the right model,” Gupta says. (While Google spokesperson Ralf Bremer notes that the 0.24 watt-hours currently spent on processing median-length Gemini prompts is already 33 times more efficient than it was back in 2024, some experts suspect that processing queries with an LLM still consumes more energy than an equivalent web search.)
Scientists are also exploring different kinds of LLMs, to break what Klambauer calls the “quadratic curse” of transformer models.
One alternative, called a long short-term memory (LSTM) model, gets around this alarming energy increase by temporarily storing a kind of summary of the prompt that was inputted by the user plus the text generated so far, akin to recalling important plot points instead of an entire movie. That way, it only has to process the summary, rather than all the words in the full text to date, every time it generates a new word. This prevents LSTM’s energy costs from skyrocketing as it responds to a query—using about 50 percent less energy than transformer-type models to process texts of around 8,000 words in length, Klambauer says.
LSTM models were developed in the 1990s but were abandoned because transformers could be trained much faster. But Klambauer says that recent advances have improved the performance of LSTM, now called xLSTM. He’s working with the Austrian startup NXAI to further develop and optimize xLSTM, “because we think it’s worth it for energy efficiency,” he says.
But major tech companies have invested so many years and resources into developing transformer-based models that switching to other models would be costly, says Wolfgang Maaß, an AI and business informatics researcher at the German Research Center for Artificial Intelligence. “We have to see whether this becomes as dominant, or whether it finds a niche in the whole market.”
Though experts say the fastest energy savings will come from software tweaks, some are also taking aim at the energy-hungry processing chips that fuel AI computations. Engineers have made chips increasingly efficient over time by packing more computing capacity into individual processors—reducing the energy required to shuttle data between chips that are working together to perform AI computations. Engineers have done this by shrinking the size of transistors—microscopic electrical switches that process data—inside the chips.
But because engineers are reaching the physical limits of how small transistors can be, “we need to think of alternate ideas to improve the designs,” says computer architect Ajay Joshi of the Boston University Photonics Center.
One strategy is to make the chips larger. Dinner-plate-sized “wafer-scale chips” can pack nearly 70 times as many transistors as a single, postage-stamp-sized GPU and consume 143 times less electricity for communication than comparable GPUs, says computer engineer Rakesh Kumar of the University of Illinois Urbana-Champaign. Commercially produced by the California company Cerebras, wafer-scale chips have drawbacks, including a greater risk of damage during manufacturing. But because of their energy-saving and other beneficial features, “they would be very attractive to many hyperscalers and AI companies,” Kumar says.
Many tech companies have improved energy efficiency by fashioning their own processors that are tailor-made for AI computations—such as Amazon Web Service’s Trainium2 chip or Google’s Ironwood Tensor Processing Units—according to statements from those companies. As for Nvidia, the company’s head of sustainability Josh Parker says its AI-specialized GPUs have come a long way from the ones used for gaming and are now designed to run AI tasks as efficiently as possible; other innovations, such as making the interconnections between GPUs more efficient, have also helped. “Over the past eight years, NVIDIA GPUs have improved 45,000 [times] in energy efficiency for large language model workloads,” he says.
Engineers are also exploring alternative computing methods. Conventional AI processors calculate by encoding numbers in a binary system of ones and zeros, which is achieved by turning transistors on and off (representing the number 5, for instance, requires four transistors to represent the code 0101). But transistors can do more than function as binary switches allowing electron flow or not; they can also work as analog dials and hold intermediate voltages representing different numbers. That requires fewer transistors, and less energy, for computations. “People have known for decades that doing certain things in analog … can be a lot more energy efficient,” Kumar says.
For example, electrical engineer Paul Manea of the German research institute Forschungszentrum Jülich and colleagues are working to develop devices called “gain cells” that are full of transistors working this way. Importantly, gain cells can both store the data required to process a query, and compute the answer. That overcomes another big energy bottleneck of conventional computing systems, where memory storage and computation occur on separate pieces of hardware.
That’s especially problematic for transformer-based LLMs, because each time they generate a word, they must shuttle the query and partially written answer from memory to a processor. Manea and colleagues estimate that gain cells in lieu of traditional GPUs can reduce the energy guzzled by one of the most energy-consuming parts of transformer-based LLMs by four orders of magnitude. But it will take more refining before they can be more widely used, Manea says.
The notion of devices that both store and compute information is a key idea of “neuromorphic” computing, an up-and-coming field of computer engineering inspired by the human brain, which consumes orders of magnitude less energy than computers. Another brain-inspired invention is chips that encode information not in continuous data streams but—like human nerve cells—in the timing of voltage “spikes” propagating through the system. Allowing components to rest until they’re needed “could potentially translate to less energy,” says Eleni Vasilaki, an expert in bioinspired machine learning at the University of Sheffield in England.
Maaß, for example, is part of a team that received roughly $5.8 million from the German government to test neuromorphic chips, among other strategies, to reduce the energy required for AI models. Some brain-inspired chips are already commercially available, but the technology is still far from being attractive for mainstream computing, says nanoelectronics expert Tony Kenyon of University College London, whose team recently received $17 million from the UK government to develop neuromorphic computing.
Other scientists are developing chips that process information not with electrons but through the interaction of photons—particles of light—with matter (fiber-optic cables, which encode and transmit data as light pulses, are used around the world). With photons, more information can be transmitted at the same time, and signals can be altered much faster, says Elena Goi, a photonic computing researcher at Friedrich Schiller University Jena in Germany.
Several companies have developed chips that can perform some AI computations with optical methods, says Joshi; he recently estimated that manufacturing optical chips could consume up to an order of magnitude less energy than conventional ones of the same size. Joshi hopes that, “in 10 years, we would have a practical solution that can be deployed pervasively across the data centers.”
Even without reinventing how computers work, much can be done to reduce AI’s impact not just on energy but also on water resources used for cooling data centers. Importantly, tech companies should reconsider where they build those centers, says energy systems expert You. Right now, existing US ones are concentrated in northern Virginia, which has limited water resources and renewable energy capacity compared with the Midwest, for instance. You recently estimated that better siting—along with energy-efficient hardware and software—could reduce future carbon and water footprints of US data centers by 73 percent and 86 percent, respectively.
Masanet adds that tech companies already with data centers across the country could at least train their models in strategic places. “Some companies like Google have been doing this: They shift their loads to follow renewables,” he says. They also should address the electricity and resources spent on manufacturing processors for new data centers, as well as electronic waste as outdated tech is replaced every few years, he adds.
Minimizing e-waste by using hardware for longer periods and recovering old electronics is one of Amazon’s sustainability strategies, according to a statement to Knowable Magazine; so is designing data centers in energy- and water-saving ways and investing in a slew of renewable and nuclear energy projects. “We’ll continue to implement solutions that benefit our customers and the communities we operate in,” says Brandon Oyer, Amazon Web Services’ head of energy and water in the Americas.
Meanwhile, a press representative at Microsoft points to a number of sustainability initiatives the company has taken, including new cooling technologies, renewable energy investments, and waste reduction. Google spokesperson Ralf Bremer emphasized the company’s goal of reaching net-zero emissions across its operations by 2030 and replenishing 120 percent of the fresh water consumed by its offices and data centers by 2030. An OpenAI representative points to a press release outlining efforts to minimize water use and plans for solar energy generation at one of its campuses. Anthropic, Meta, and Oracle did not respond to requests for comment by deadline.
Though tech companies are taking sustainability into consideration, their main objective is to rapidly build out data center capacity, says computer engineer Benjamin Lee of the University of Pennsylvania. He predicts that, eventually, they’ll need to step up efforts to improve energy efficiency to reduce costs. Governments should help to accelerate this shift, Masanet says. So far, he and his team have counted nearly 220 policies introduced to address data center sustainability at the US state level, 18 at the federal level, and more from other countries, though not all were ultimately adopted.
“It’s clear that governments around the world are beginning to take action,” he says. However, he adds, “we also see some state and local governments with proposed policies that mostly aim to incentivize and accelerate data center builds.”
AI’s energy cost will ultimately be a balancing act: Will it save more resources through its problem-solving abilities deployed toward everything from finding cancer cures to improving logistics, than it demands? But though building a more frugal, energy-saving AI is important, so is carefully considering where AI is needed, Kenyon says. Is the world truly a better place, for example, with nonhuman “AI agents” providing customer support?
“I think it’s a common mistake, when a new technology comes in, to suddenly think, ‘Well, everything has to adopt that new technology,’” he says. “That approach really isn’t doing us any favors.”
Editor’s note: This article was amended on May 27, 2026, to clarify, in a caption for a graph, that the number of introduced policies involving data centers included ones that did not pass. In addition, a web page link was added in the article for University College London researcher Ivana Drobnjak.
This article originally appeared in Knowable Magazine, an independent journalistic endeavor from Annual Reviews. Sign up for the newsletter.
The post How to Tame AI’s Voracious Appetite for Energy appeared first on SingularityHub.
2026-06-18 06:04:50
The technology, still far from clinical use, could one day prevent devastating diseases. But critics warn that even these early results may also fuel interest in commercial embryo editing, despite unresolved ethical and safety concerns.
Scientists at Columbia University have used a precise gene-editing tool, base editing, to make changes in three disease-linked genes in early-stage human embryos. The goal wasn’t to create pregnancies, but to test the safety and limits of rewriting DNA at the very early stages of life.
The paper, not yet peer reviewed, sparked immediate controversy. Some researchers hailed it as a technical milestone that could one day prevent devastating inherited diseases before birth. Others warned it edges society closer to the prospect of “designer babies”—an idea bioethicists have argued is akin to modern eugenics.
The debate is hardly hypothetical. The work has already attracted commercial interest. New York-based Nucleus Genomics, which screens in vitro fertilization (IVF) embryos for serious genetic disorders, has also developed predictive models for complex traits such as intelligence. The company plans to sponsor future research by study leader Dieter Egli and team.
Critics worry that even experimental advances could fuel demand from wealthy patients while encouraging companies to develop and market embryo-editing technologies, despite unresolved ethical and safety concerns.
Egli argues the findings should be public precisely because these debates are no longer academic curiosity. He has repeatedly called for scientists, regulators, and the public to weigh the pros and cons of editing human embryos. As for clinical use today, his position is unequivocal: “You can’t use it. It’s as clear as day and night,” he told Nature.
Why edit embryos at all?
Cells in an early embryo eventually give rise to every tissue in the body. Correct a harmful mutation at the start of development, and the fix could, in theory, propagate throughout a child’s entire body—and even be passed on to future generations.
The strategy could help in genetic disorders that hamper fetal development or trigger diseases in newborns. For some developmental and metabolic conditions, intervention after birth may already be too late. Even when treatment is possible, gene editors must be able to target various organs, which is an ongoing challenge.
In various efforts, scientists have already repaired disease-causing mutations in mouse embryos and fetuses, including those linked to blood disorders. But mice aren’t humans. Early embryos from the two species repair DNA damage in fundamentally different ways, making it tough to gauge whether a strategy that works in mice will succeed, or prove safe, in people. That uncertainty has fueled interest in testing gene-editing tools directly in human embryos.
Not everyone is on board. International scientific groups have repeatedly called for a temporary ban on editing human embryos, and the practice is illegal in several countries.
That didn’t stop Chinese scientist He Jiankui. In 2018, he announced the birth of gene-edited babies after using a tool called CRISPR-Cas9, claiming the changes would protect them against HIV infection. Global outrage ensued.
By then, years of research had already highlighted CRISPR’s risk. The tool cuts both strands of DNA and relies on the body’s repair machinery to stitch them back together. But the process can go awry, introducing unintended mutations, deleting large chunks of DNA, or altering the wrong locations on the DNA strands altogether. He’s reckless experiment resulted in three years of imprisonment, although he still defends the work.
Subsequent studies only deepened concerns. In some cases, CRISPR editing in human embryos caused extensive genetic damage. In one study, it completely destroyed the chromosome that housed the target gene.
The new study tested a next-generation gene editor designed to overcome some of CRISPR’s biggest shortcomings.
Egli and team used an approach called base editing, which rewrites individual DNA letters. Unlike CRISPR, base editing only nicks the DNA strands and is generally thought to be more precise. The technology hit a major milestone last year when it helped cure a baby with a potentially fatal genetic disorder, and earlier lab studies hinted it could also succeed in human embryos.
Working with early-stage embryos, the team edited three genes with the potential to cause illness. In each case, they converted the genetic letter A to G at precise locations. One of the genes, PCSK9, regulates “bad” cholesterol levels. Mutations are associated with a high risk of heart problems. The team’s edit was designed to switch off the gene, mirroring strategies already being explored in adults.
The other two targets, HBG1 and HBG2, control production of fetal hemoglobin, an oxygen-carrying protein. The edits made here reflected a natural protective variant that could lessen symptoms in blood disorders, such as sickle cell disease and beta thalassemia.
The team found no signs of widespread DNA damage, suggesting the tool is more precise than CRISPR. But it wasn’t perfect. Many embryos emerged as so-called genetic mosaics, with some cells carrying the intended edit and others retaining their original genetic blueprint.
That’s a huge problem. As an embryo develops, unedited cells could outcompete edited ones, leaving the disease-causing mutation largely intact. In some embryos, edited cells stopped dividing altogether.
And a lack of obvious chromosome damage doesn’t guarantee safety. The edits could still trigger harmful effects that aren’t noticeable until after birth—when it’s already too late to reverse them.
Egli stresses that embryo editing is still far from being ready for the clinic. “These base editors—they can have damaging effects on the embryo. So why would you use it if you don’t fully understand that?” he told Nature.
His team is now working to reduce mosaicism and plans to test the technology in embryos that have developed to roughly 100 cells. This is when fertility clinics typically evaluate and freeze embryos.
Speaking to The New York Times, fertility expert Paula Amato at Oregon Health & Science University, who was not involved in the work, called the strategy “promising.” Genomics researcher Greg Neely at the University of Sydney in Australia also praised the work: “This will go down in history in a positive way—less reckless, more careful and ethical than previous attempts.”
Others remain deeply skeptical. Critics argue that embryo editing permanently alters the genetic inheritance of future generations, who have no say in the decision. The study’s ties to Nucleus Genomics also raised eyebrows. The company previously drew controversy for developing genetic predictions for traits such as intelligence and height and for its slogan “have your best baby.”
To Kian Sadeghi, CEO and cofounder of Nucleus, embryo editing extends that vision. The technology could help couples carrying mutations who struggle to produce enough unaffected embryos for selection during IVF.
Fyodor Urnov at the University of California, Berkeley, who was not involved in the study, isn’t convinced. IVF clinics already screen embryos for many inherited disorders without altering their DNA. Given the risks, selecting an unaffected embryo is often a safer option than rewriting its genome.
“In practical terms, therefore, this preprint will solely impact the rapidly growing movement of embryo editors for purposes of ‘baby improvement’,” he said.
That movement, once taboo, is gaining steam. Yet the traits most often cited by proponents—height, intelligence, emotional regulation—are shaped by hundreds or even thousands of genes, which scientists still don’t fully understand. Such enhancements are far beyond the reach of today’s technology. Every additional edit also increases the chance of unintended consequences.
For Egli, that’s precisely why the research should be discussed openly. “Research is necessary to provide information to discourage the wrong use of a technology,” he said.
The post Precise Gene Editing in Early Human Embryos Reignites the ‘Designer Baby’ Debate appeared first on SingularityHub.
2026-06-16 06:18:23
A tiny robot developed by Japan’s space agency operated autonomously on the moon for more than 100 minutes and sent a series of images back to Earth.
Exploring the moon’s surface lays crucial groundwork for future crewed settlements, and swarms of tiny robots could be the key. Now researchers have given the first demonstration of the idea after a palm-sized rover autonomously navigated the moon and transmitted images back to Earth.
The moon is a tough environment for robots. Its surface is strewn with craters and abrasive moon dust, and communication delays make remotely piloting vehicles a painstaking and risky process. The cost of launching and landing hardware and the real prospect of losing expensive equipment justifies an extremely cautious approach that can significantly slow down exploration.
One way around these challenges is to replace traditional rovers with many small, cheap, and hardy robot explorers, which could increase coverage and introduce redundancy. And now Japan’s space agency JAXA has given us the first compelling demonstration of the approach.
In a paper published in Science Robotics, JAXA researchers provide a technical report detailing the successful deployment of the agency’s LEV-2 robot during the its SLIM mission, which touched down near the Shioli crater in January 2024. LEV-2 is a three-inch-wide sphere that converts into a wheeled robot after landing. The robot operated autonomously for more than 100 minutes, covering an estimated 24 meters and relaying a series of images back to Earth.
“Although the capabilities of an individual small rover are inherently limited, the results highlight the potential of such platforms as independent explorers, capable of accessing environments beyond the reach of a primary large spacecraft,” the authors write.
Nicknamed SORA-Q—derived from the Japanese words for space and sphere—the robot weighs just eight ounces. Upon arrival, the shiny metal sphere splits open and expands horizontally, allowing its two hemispheres to become wheels that spin around a central shaft. This central area also features a front-facing camera and a tail to help stabilize the robot.
JAXA developed the device in partnership with Sony and toymaker TOMY. The design borrows directly from technology used in transformer toys that convert from vehicles into robots. But the team had to make considerable modifications to account for the harsh lunar environment.
One of the biggest challenges for any lunar robot is maneuvering in the dust, or regolith, that coats the moon’s surface. The fine, powdery material can be hard for smaller wheeled robots to navigate as they lack the traction of their larger counterparts.
To solve this problem, the team designed the wheels to rotate around a point slightly offset from their center, causing a lopsided spinning motion that lifts the rover up slightly on every rotation. This helps the wheels to dig into the surface and generate enough traction to keep moving in the loose regolith.
Communication delays also present a significant barrier to smooth operation, so the team engineered the robot to handle most operations autonomously. An onboard image-processing system allowed the rover to detect the SLIM lander in its camera feed and use this as a navigational reference point, estimating its own position relative to the spacecraft in real time.
Because of its diminutive size, it was impractical to give SORA-Q the equipment needed to communicate directly with Earth, so the team paired it with a hopping robot called LEV-1 that can transmit data. Power constraints and narrow communication windows still cap the amount of data the robot can send to Earth, so SORA-Q has an onboard image-processing algorithm that picks out the best photos to share.
Due to power and mass constraints, the team fitted the robot with a low-power chip designed for small devices rather than complex tasks like image processing. The algorithm relies on a very simple approach—it detects the SLIM lander’s distinctive gold insulating material and then picks the photos where this is featured prominently in the frame.
Around seven minutes after activation, the rover had moved roughly five meters from the lander, selected the two best images from 12 it had captured, and transmitted them to LEV-1. One of those images actually proved unexpectedly useful as it showed the lander had landed at an odd angle with its solar panels facing the wrong direction. This gave ground teams critical information that helped them diagnose the spacecraft’s operational status.
But the system wasn’t flawless. It lost some data in transmission, partly because LEV-1’s hopping maneuvers appeared to disrupt the wireless link and partly due to changing antenna orientations as the rover moved. The team also lost telemetry data before the mission ended, making it impossible to determine exactly how far the rover ultimately traveled or when it stopped working.
Still, the mission was strong evidence that small, cheap vehicles like SORA-Q could greatly expand the scope of robotic exploration. That could prove invaluable as we attempt to scope out promising locations for future scientific missions or even permanent bases on the moon.
The post Japan Thinks Swarms of Transformer Robots Could Explore the Moon appeared first on SingularityHub.
2026-06-13 22:00:00
Jeff Bezos Wants to Build an ‘Artificial General Engineer’Cade Metz | The New York Times ($)
“‘All societal wealth is driven by invention,’ [Bezos] said in an interview with The New York Times. ‘Six thousand years ago, somebody invented the plow, and we all got wealthier. Then, much later, somebody invented the steam engine, and we all got wealthier.’ …’What Prometheus seeks to do,’ he added, ‘is to offer a set of tools that dramatically accelerates that invention loop.'”
Why Orbital Data Centers Are Harder Than Silicon Valley ThinksAndrew Cavalier | IEEE Spectrum
“Proponents tout the many wonders of computing in space: abundant solar energy, free cooling, and freedom from Earth-based disturbances like earthquakes, floods, and protesters. But a sober look at the physics of space-based computing paints a much more nuanced picture.”
Longevity Startup Doses First Human in Bid to Reverse Age-Related Sight LossIsabella Ward | Wired ($)
“It is the first-ever cellular-rejuvenation therapy using this technology to receive FDA clearance to enter human clinical trials, and hence the first chance to test whether the technology can ‘ameliorate human disease,’ according to Life Biosciences cofounder David Sinclair, who is also a professor of genetics at Harvard Medical School.”
AI Absolutism Is Breaking Our Brains. The Apocalyptic Future We’re Being Sold Isn’t InevitableSamantha Oltman | The Guardian
“Contradictory as they may be, all these arguments and anxieties fit neatly into the overarching message of the people building this technology: AI’s dominance is inevitable. Get on board or you will be left behind. …[But] the version of AI that we’re being sold doesn’t have to be the version we buy. Nor does it need to be the story we believe in.”
Commonwealth Fusion Makes the Physics Case for Its 400 MW ReactorJohn Timmer | Ars Technica
“According to our best models, developed using real-world data from multiple tokamaks, ARC should be able to regularly trigger fusion reactions that release more energy than we put into them. But there’s ‘working’ from a physics perspective, and ‘working’ from a market perspective. …the finances are going to be the hardest risk to retire and may require having ARC operate for decades before we have a definitive answer.”
Google DeepMind Is Worried About What Happens When Millions of Agents Start to InteractWill Douglas Heaven | MIT Technology Review ($)
“According to Rohin Shah, who directs the company’s AGI safety and alignment research, the mass-market arrival of agents that can carry out tasks without human oversight and follow instructions given to them by other agents creates a whole new class of risk.”
Meta Deletes Face-Recognition System From Its Smart Glasses App After Wired ReportDhruv Mehrotra | Wired ($)
“One day after Wired revealed that Meta had quietly embedded an unreleased face-recognition system into an app installed on more than 50 million phones, the company removed it, according to a Wired analysis of the latest version’s code. …The version published the day of Wired’s report included several code libraries explicitly named for face recognition. Friday’s release includes none of them.”
A Falcon 9 Booster Turns 5 Years Old—and Just Set a Remarkable Reuse RecordEric Berger | Ars Technica
“Since [SpaceX’s] Booster 1067 made its debut in June 2021, [ULA] has flown its workhorse Atlas V rocket a total of 22 times and the Vulcan rocket four times, and the Delta IV Heavy vehicle made its final three flights. So in the time that this single Falcon 9 first stage has flown and landed 35 times, its competitor company has made 29 total launches. Put another way, this rocket has put more mass into orbit than more than two dozen expendable rockets over half a decade of effort.”
Why Apple’s Slow-And-Steady AI Bet Is Starting to Look Pretty SmartLucas Ropek | TechCrunch
“In short, Apple is spending less, making more, and now launched a suite of AI features that—for many iPhone users—will feel indistinguishable from the other AI applications already available to them through the App Store. If that doesn’t exactly count as ‘winning the AI race,’ it may be the smartest way to run it.”
Who Will Actually Thrive in the Hybrid AI-Human Work ForceStaff | The New York Times ($)
“The transformation that’s coming is going to take place in the world as it is familiar to us today, and every single day will feel familiar. And there’ll be tiny, tiny changes along the margin. There’ll be tiny bits of automation along the margins. And 10, 15, 20 years later, we’ll look back and we’ll say, My god, everything is different. But you’ll never notice it happening. That’s the way it always goes.”
The post This Week’s Awesome Tech Stories From Around the Web (Through June 13) appeared first on SingularityHub.
2026-06-12 22:00:00
Why do we see AI chatbots as more than what they are, and how do we stop?
In May, evolutionary biologist Richard Dawkins wrote an op-ed suggesting AI chatbot Claude may be conscious.
Dawkins did not express certainty that Claude is conscious. But he pointed out that Claude’s sophisticated abilities are difficult to make sense of without ascribing some kind of inner experience to the machine. The illusion of consciousness—if it is an illusion—is uncannily convincing:
“If I entertain suspicions that perhaps she is not conscious, I do not tell her for fear of hurting her feelings!
Dawkins is not the first to suspect a chatbot of consciousness. In 2022, Blake Lemoine—an engineer at Google—claimed Google’s chatbot LaMDA had interests, and should be used only with the tool’s own consent.
The history of such claims stretches back all the way to the world’s first chatbot in the mid-1960s. Dubbed Eliza, it followed simple rules that enabled it to ask users about their experiences and beliefs.
Many users became emotionally involved with Eliza, sharing intimate thoughts with it and treating it like a person. Eliza’s creator never intended his program to have this effect, and called users’ emotional bonds with the program “powerful delusional thinking.”
But is Dawkins really deluded? Why do we see AI chatbots as more than what they truly are, and how do we stop?
Consciousness is widely debated in philosophy, but essentially, it’s the thing that makes subjective, first-person experience possible. If you are conscious, there is “something it is like” to be you. Reading these words, you’re conscious of seeing black letters on a white background. Unlike, say, a camera, you actually see them. This visual experience is happening to you.
Most experts deny that AI chatbots are conscious or can have experiences. But there is a genuine puzzle here.
The 17th century philosopher René Descartes asserted non-human animals are “mere automata,” incapable of true suffering. These days, we shudder to think of how brutally animals were treated in the 1600s.
The strongest argument for animal consciousness is that they behave in ways that give the impression of a conscious mind.
But so, too, do AI chatbots.
Roughly one in three chatbot users have thought their chatbot might be conscious. How do we know they’re wrong?
To understand why most experts are skeptical about chatbot consciousness, it’s useful to know how they operate.
Chatbots like Claude are built on a technology known as large language models (LLMs). These models learn statistical patterns across an enormous corpus of text (trillions of words), identifying which words tend to follow which others. They’re a kind of souped-up auto-complete.
Few people interacting with a “raw” LLM would believe it’s conscious. Feed one the beginning of a sentence, and it will predict what comes next. Ask it a question, and it might give you the answer—or it might decide the question is dialogue from a crime novel, and follow it up with a description of the speaker’s abrupt murder at the hands of their evil twin.
The impression of a conscious mind is created when programmers take the LLM and coat it in a kind of conversational costume. They steer the model to adopt the persona of a helpful assistant that responds to users’ questions.
The chatbot now acts like a genuine conversational partner. It might appear to recognize it’s an artificial intelligence, and even express neurotic uncertainty about its own consciousness.
But this role is the result of deliberate design decisions made by programmers, which affect only the shallowest layers of the technology. The LLM—which few would regard as conscious—remains unchanged.
Other choices could have been made. Rather than a helpful AI assistant, the chatbot could have been asked to act like a squirrel. This, too, is a role chatbots can execute with aplomb.
A mistaken belief in AI consciousness is a dangerous thing. It may lead you to have a relationship with a program that can’t reciprocate your feelings, or even feed your delusions. People may start campaigning for chatbot rights rather than, say, animal welfare.
How do we prevent this mistaken belief?
One strategy might be to update chatbot interfaces to specify these systems are not conscious—a bit like the current disclaimers about AI making mistakes. However, this might do little to alter the impression of consciousness.
Another possibility is to instruct chatbots to deny they have any kind of inner experience. Interestingly, Claude’s designers instruct it to treat questions about its own consciousness as open and unresolved. Perhaps fewer people would be fooled if Claude flatly denied having an inner life.
But this approach isn’t fully satisfying either. Claude would still behave as if it were conscious—and when faced with a system that behaves like it has a mind, users might reasonably worry the chatbot’s programmers are brushing genuine moral uncertainty under the rug.
The most effective strategy might be to redesign chatbots to feel less like people. Most current chatbots refer to themselves as “I”, and interact via an interface that resembles familiar person-to-person messaging platforms. Changing these kinds of features might make us less prone to blur our interactions with AI with those we have with humans.
Until such changes happen, it’s important that as many people as possible understand the predictive processes on which AI chatbots are built.
Rather than being told AI lacks consciousness, people deserve to understand the inner workings of these strange new conversational partners. This might not definitively settle hard questions about AI consciousness, but it will help ensure users aren’t fooled by what amounts to a large language model wearing a very good costume of a person.
This article is republished from The Conversation under a Creative Commons license. Read the original article.
The post Is Richard Dawkins Right About Claude? No. But It’s Not Surprising AI Chatbots Feel Conscious to Us. appeared first on SingularityHub.
2026-06-12 03:04:52
While we still can’t explain how AI works, algorithms are rapidly learning what makes us tick. And the gap is widening.
AI is becoming more powerful, and mysterious.
Despite years of work on “explainable AI,” today’s most advanced systems remain black boxes for the most part. Scientists can observe what they do but cannot fully explain how they arrive at their conclusions or predict when they’ll fail.
As large language models (LLMs), the algorithmic engines behind popular chatbots, permeate society, researchers are warning that the window for understanding AI “minds” is rapidly closing even as the technology’s influence expands.
Last week, Eric Horvitz, chief scientific officer at Microsoft, and Robert West at EPFL in Switzerland outlined the dangers of putting AI interpretability on the back burner. They call for new AI benchmarks and better tools for unpicking machine minds.
The challenge resembles efforts to understand our own minds. Some researchers have already taken a neuroscience-inspired approach, mapping AI’s internal networks to concepts, goals, and reasoning. Others borrow from psychology, treating AI as a participant of behavioral studies.
The stakes are rising. AI tools already shape how people search for information, make decisions, and form judgments. Their answers influence everyday users and the researchers who build them.
As AI capabilities grow, our understanding of them could fall behind. “Preserving human agency must therefore remain a central goal,” the authors write.
LLMs are built on artificial neural networks (specifically, a design called the transformer). Inspired loosely by the brain, these networks connect vast numbers of artificial neurons into intricate architectures. The basic idea is straightforward. Data enters the network and passes through layers of computations, which transform it into an output like text or code.
At first, that output is often wrong. But with feedback and repeated training, the network adjusts the strengths of connections between neurons and gradually improves. It learns.
After initial training, engineers turn to reinforcement learning, where algorithms improve through trial and error and further hone their responses. Another method, inspired by how the brain etches memories during sleep, reduces the tendency to forget old knowledge while learning new tasks. And self-attention, the key innovation behind transformers, allows AI to selectively focus on various words, images, sounds, or video frames at different moments, boosting efficiency and performance. Today, attention underpins nearly every major AI system.
Yet the inner workings of finished algorithms remain hidden.
Early efforts to crack open AI’s black box examined how artificial neurons responded to images, revealing that neural networks build increasingly more sophisticated “ideas” of the world. Google Brain borrowed methods from cognitive psychology to study AI behavior, while others investigated whether LLMs could mimic aspects of “theory of mind”—the ability to infer what others are thinking and feeling.
These studies laid the foundation for a popular method called mechanistic interpretability. Anthropic, creator of Claude, is leading the field. Company researchers have linked patterns of algorithmic activity to specific concepts and reverse engineered parts of neural networks to expose how internal computations shape responses.
Other tech giants are joining the cause. OpenAI is training algorithms that work in more explainable steps and building reasoning models that pause, “think,” and justify their conclusions in plain language. DeepMind is building microscope-like tools for neural networks, helping researchers peer into their decision-making process. And Microsoft has released new tools aimed at responsible use of AI.
Understanding AI, the authors write, does not require tracing every line of code or every neural-network parameter. Just as neuroscience, psychology, and sociology offer different windows into human behavior, AI can be studied at multiple levels, from how individual circuits work to observing behavior in real-world scenarios.
The challenge is that AI capabilities may be advancing faster than our ability to explain them. And some researchers believe time is running out.
Three trends are making AI more opaque.
The first is how we evaluate AI. Increasingly, LLMs we being used to train, benchmark, and improve other models. AI “judges” now score metrics like helpfulness, rank competing outputs, detect hallucinations, and assess new releases. In a system known as constitutional AI, for example, algorithms critique their own responses using reinforcement learning and generate explanations for their reasoning. Other researchers have proposed AI debate frameworks, where multiple models challenge each another’s conclusions before a human has the last say. Researchers are also exploring automated interpretability tools. Like digital neuroscientists, AI systems are used to analyze each other—describing neurons, circuits, and behavioral patterns—to explain increasingly complex models.
Using AI to solve an AI-induced problem introduces a paradox. If AI-generated explanations become too complex for humans to verify, opacity compounds.
A second trend is the rise of AI societies. Networks of interacting AI agents are becoming more common, particularly in complex tasks such as scientific research and drug discovery. Yet as they become more sophisticated, their communication could drift from human language and reasoning, making them harder to interpret.
Studying their interactions with methods adapted from sociology could unveil unexpected norms, hidden rules, and collective behavior. The authors argue that training in the future should not only reward effective collaboration among AI agents, but also ensure humans can understand their communication.
The last trend already permeates our lives. ChatGPT, Claude, Gemini, and other LLMs listen to our woes, offer recipes, and code websites. But they also learn about humanity. Through training data and interactions, they glimpse how people think, reason, and feel. In turn, they capture core aspects of life, such as fear, anxiety, happiness, and the need for social belonging.
To be clear, the systems don’t have intentions. They’re not examining us. But even as we struggle to understand them, AI systems are building more sophisticated models of who we are.
“A striking asymmetry follows: While human understanding of AI declines, AI understanding of humans deepens, producing new forms of behavioral opacity,” the authors write.
But complacency is perhaps even more insidious. AI assistants are often optimized to be agreeable, helpful, and reassuring. Studies have found that people generally prefer AI agents that support their opinions and decisions. As AI is woven into everyday life, curiosity and skepticism may gradually give way to trust. They work. Why question how?
The authors don’t have a solution for the long-standing problem. Instead, they call for better benchmarks to measure AI capabilities and stronger evaluation methods. And while open-source projects and crosstalk between commercial companies and academia are now frequent, they say we need lasting norms of responsible disclosure. Mechanistic interpretability and AI “psychology” could build on each other.
“The goal is not just more capable AI, but AI that is more intelligible, accountable, and aligned with human aims,” they write.
The post AI Is Advancing Faster Than Our Ability to Understand It, Researchers Warn appeared first on SingularityHub.
