By Thomas Maxwell Published March 6, 2025 | Comments (1) | Google co-founder Eric Schmidt, others say U.S. is approaching Cold War-style arms race in AI defense technology. South Korean Defense Ministry/Getty Former Google CEO Eric Schmidt and Scale AI founder Alexandr Wang are co-authors on a new paper called Superintelligence Strategy that warns against the U.S. government creating a Manhattan Project for so-called Artificial General Intelligence (AGI) because it could quickly get out of control around the world. The gist of the argument is that the creation of such a program would lead to retaliation or sabotage by adversaries as countries race to have the most powerful AI capabilities on the battlefield. Instead, the U.S. should focus on developing methods like cyberattacks that could disable threatening AI projects. Schmidt and Wang are big boosters of AIs potential to advance society through applications like drug development and workplace efficiency. Governments, meanwhile, see it as the next frontier in defense, and the two industry leaders are essentially concerned that countries are going to end up in a race to create weapons with increasingly dangerous potential. Similar to how international agreements have reined in the development of nuclear weapons, Schmidt and Wang believe nation states should go slow on AI development and not fall prey to racing one another in AI-powered killing machines. At the same time, however, both Schmidt and Wang are building AI products for the defense sector. The formers White Stork is building autonomous drone technologies, while Wangs Scale AI this week signed a contract with the Department of Defense to create AI agents that can assist with military planning and operations. After years of shying away from selling technology that could be used in warfare, Silicon Valley is now patriotically lining up to collect lucrative defense contracts.All military defense contractors have a conflict of interest to promote kinetic warfare, even when not morally justified. Other countries have their own military industrial complexes, the thinking goes, so the U.S. needs to maintain one too. But in the end, innocent people suffer and die while powerful people play chess. Palmer Luckey, the founder of defense tech darling Anduril, has argued that AI-powered targeted drone strikes are safer than launching nukes that could have a larger impact zone or planting land mines that have no targeting. And if other countries are going to continue building AI weapons, we should have the same capabilities as deterrence. Anduril has been supplying Ukraine with drones that can target and attack Russian military equipment over enemy lines.Anduril recently ran an ad campaign that displayed the basic text Work at Anduril.com covered with the word Dont written in giant, graffiti-style spray-painted letters, seemingly playing to the idea that working for the military industrial complex is the counterculture now. Schmidt and Wang have argued that humans should always remain in the loop on any AI-assisted decision making. But as recent reporting has demonstrated, the Israeli military is already relying on faulty AI programs to make lethal decisions. Drones have long been a divisive topic, as critics say that soldiers are more complacent when they are not directly in the line of fire or do not see the consequences of their actions first-hand. Image recognition AI is notorious for making mistakes, and we are quickly heading to a point where killer drones will fly back and forth hitting imprecise targets. The Schmidt and Wang paper makes a lot of assumptions that AI is soon going to be superintelligent, capable of performing as good if not better as humans in most tasks. That is a big assumption as the most cutting-edge thinking models continue to produce major gaffs, and companies get flooded with poorly-written job applications assisted by AI. These models are crude imitations of humans with often unpredictable and strange behavior. Schmidt and Wang are selling a vision of the world and their solutions. If AI is going to be all-powerful and dangerous, governments should go to them and buy their products because they are the responsible actors. In the same vein, OpenAIs Sam Altman has been criticized for making lofty claims about the risks of AI, which some say is an attempt to influence policy in Washington and capture power. It is sort of like saying, AI is so powerful it can destroy the world, but we have a safe version we are happy to sell you.Schmidts warnings are not likely to have much impact as President Trump drops Biden-era guidelines around AI safety and pushes the U.S. to become a dominant force in AI. Last November, a Congressional commission proposed the Manhattan Project for AI that Schmidt is warning about and as people like Sam Altman and Elon Musk gain greater influence in Washington, its easy to see it gaining traction. If that continues, the paper warns, countries like China might retaliate in ways such as intentionally degrading models or attacking physical infrastructure. It is not an unheard of threat, as China has wormed its way into major U.S. tech companies like Microsoft, and others like Russia are reportedly using freighter ships to strike undersea fiber optic cables. Of course, we would do the same to them. Its all mutual. It is unclear how the world could come to any agreement to stop playing with these weapons. In that sense, the idea of sabotaging AI projects to defend against them might be a good thing.Daily NewsletterYou May Also Like By Isaac Schultz Published March 5, 2025 By Matt Novak Published March 5, 2025 By Thomas Maxwell Published March 4, 2025 By AJ Dellinger Published February 28, 2025 By Thomas Maxwell Published February 28, 2025 By Thomas Maxwell Published February 27, 2025

Star Wars Rebels is about to come back into our lives in a very fun way and we cant wait. Rebels, of course, was the brainchild of Dave Filoni, the man personally hired by George Lucas to help him with an earlier animated series, The Clone Wars. And when Lucas hired Filoni, he knew what he was doing. Filoni took Clone Wars and made a fantastic, multi-season exploration into the mythical Star Wars event while also giving fans a deeper appreciation for iconic characters like Obi-Wan Kenobi and Anakin Skywalker. Then, when Clone Wars ended, Filoni followed it up with something even more ambitious. A show that, at least to start, would feature all new characters in a much less defined timeline. Meaning hed have to build and develop the characters and stories without as many preconceptions. That show, Star Wars Rebels, ran for just four seasonsbut to us, it eclipsed Clone Wars in multiple ways. The new characters such as Ezra Bridger, Hera Syndulla, Sabine Wren, Kanan Jarrus, and Zeb Orrelios became instant fan favorites, and their adventures eventually interweaved them with Star Wars icons like Luke Skywalker, Princess Leia, Darth Vader, and, crucially, Ahsoka Tano. The story then continued in live-action with Ahsoka, on Disney+. Even though it was ostensibly a kids show, under Filonis guidance, Rebels became much, much more. And now, seven years after its conclusion, its coming back in a whole new medium. Earlier this week, news broke that three of Rebels main voice actorsVanessa Marshall, who played Hera; Tiya Sircar, who played Sabine; and Taylor Gray, who played Ezraare about to start a rewatch podcast called Pod of Rebellion. The Ghost trio will break down each episode of Rebels with the help of host Jon Lee Brody. Heres a little announcement video. As revealed there, the show starts March 11, and you can get more information on I Heart as well as Apple. This is fantastic news for Rebels fans but we already have a few questions. Could Filoni appear at some point? What about Freddie Prinze Jr., the voice of Kanan? And, will Marshall, Sircar, and Gray bring the show on the road to Star Wars Celebration? Whatever the case, well anxiously await to see what the trio are able to reveal about their fantastic show. Want more io9 news? Check out when to expect the latest Marvel, Star Wars, and Star Trek releases, whats next for the DC Universe on film and TV, and everything you need to know about the future of Doctor Who.

Dogs carrying the genetic variant most associated with obesity, DENND1B, had around 8% more body fat than those without it. (Image Credit: University of Cambridge)NewsletterSign up for our email newsletter for the latest science newsIf youre a dog lover and owner, you probably share a lot with your dog. You may share your thoughts about the day, share in the pleasure of a good, long walk, or even share a bed. Along with that, you may also share an obesity gene if you own a lab.A new study published in Science found that British Labrador retrievers have multiple genes related to canine obesity, of which are also associated with human obesity.Obesity Genes and Brain PathwaysResearchers from the University of Cambridge analyzed different genes within British Labrador retrievers. They found that a gene called DENND1B was mostly associated with obesity in the dogs. Its also a gene humans carry and is also associated with obesity.The research team noted that DENND1B directly impacts the brain pathway that regulates the energy balance in the body. This is known as the leptin melanocortin pathway. The team also noted four additional genes associated with obesity in dogs and humans, though they were not as effective as DENND1B. Researchers believe that these genes would not be a target for weight-loss drugs.These genes are not immediately obvious targets for weight-loss drugs because they control other key biological processes in the body that should not be interfered with, said Alyce McClellan from the University of Cambridges Department of Physiology, Development and Neuroscience, and co-lead author of the study, in a press release. But the results emphasise the importance of fundamental brain pathways in controlling appetite and body weight.Higher Appetite, Higher RiskThe research team analyzed the traits of the dogs with the obesity gene and noted that they all seemed to have food on the brain.We found that dogs at high genetic risk of obesity were more interested in food, said Natalie Wallis from the University of Cambridges Department of Physiology, Development and Neuroscience, and co-lead author of the study, in a press release. We measured how much dogs pestered their owners for food and whether they were fussy eaters. Dogs at high genetic risk of obesity showed signs of having higher appetite, as has also been shown for people at high genetic risk of obesity.The study authors found that dog owners who helped control how much their dogs ate and saw to it that they had regular exercise were less likely to be obese. This is similar to humans with the obesity gene; those who followed a stricter diet and exercised more often were less likely to be obese.The study authors noted that having the obesity gene does not automatically make you obese whether dog or human it just means there is a higher chance of it.The Human-Dog ConnectionAccording to the study, about 40 to 60 percent of pet dogs are obese. This could lead to a variety of different health problems, including arthritis, heart disease, and early death. This isnt too far off from the impacts obesity can have on humans.For this study, the research team surveyed pet dogs based on their greediness level and body fat and took saliva samples for DNA. They found that dogs with the DENND1B gene had 8 percent more body fat than dogs without the gene. The researchers then took this information and compared it to humans to see if there was any relevance.Studying the dogs showed us something really powerful: owners of slim dogs are not morally superior. The same is true of slim people. If you have a high genetic risk of obesity, then when theres lots of food available youre prone to overeating and gaining weight unless you put a huge effort into not doing so, said Eleanor Raffan, a researcher in the University of Cambridges Department of Physiology, Development and Neuroscience who led the study, in a press release.By studying dogs, we could measure their desire for food separately to the control owners exerted over their dogs diet and exercise. In human studies, its harder to study how genetically driven appetite requires greater willpower to remain slim, as both are affecting the one person, she added in a press release.Overall, studying the gene in dogs and humans helped the research team understand how the brain controls eating behavior and energy use.Article SourcesOur writers at Discovermagazine.com use peer-reviewed studies and high-quality sources for our articles, and our editors review for scientific accuracy and editorial standards. Review the sources used below for this article:A graduate of UW-Whitewater, Monica Cull wrote for several organizations, including one that focused on bees and the natural world, before coming to Discover Magazine. Her current work also appears on her travel blog and Common State Magazine. Her love of science came from watching PBS shows as a kid with her mom and spending too much time binging Doctor Who.1 free article leftWant More? Get unlimited access for as low as $1.99/monthSubscribeAlready a subscriber?Register or Log In1 free articleSubscribeWant more?Keep reading for as low as $1.99!SubscribeAlready a subscriber?Register or Log In

Ice may lie an inch or so below the Moons surface in many more regions than previously thought which has major implications for longer-term lunar exploration or even habitation.An India-led project found that differences in the Moons surface temperature vary widely, depending on location (poles versus equator), orientation (facing the Sun or not), and angle (diagonal crater walls versus flat surfaces). The lunar surface measurements were taken in 2023 by the Indian Chandrayaan-3 mission and were recently published in the journal Communications Earth & Environment.Taking the Moon's TemperatureThe measurements were the first to directly record the Moons surface temperature since the Apollo missions in the 1970s. Other measurements have employed remote sensing, but that data is considered both less accurate and not as effective at capturing variations based on location and geography. But unlike those missions, the Chandra mission recorded temperatures were from several spots on the moons surface at multiple times of the day, while the Apollo missions primarily took temperatures near the equator.Since the Apollo missions, others have made efforts to take the satellites temperature from different locations. But no mission, until Chandra, had been able to essentially stick a thermometer beneath the Moons surface. That may have been at least in part because previous attempts chose the wrong tools.Probing the Moon's CrustEarlier unsuccessful efforts tried to hammer probes into the Moons crust. ChaSTE, Chandras suite of data collection tools took a different approach: drilling a screw-like device into the ground. The Chandrayaan-3 mission employed a Vikram lander to reach Earths satellite and a Pragyan rover to explore its surface.To the best of our knowledge, ChaSTE is the first experiment to demonstrate penetration as a potential mechanism for deploying planetary thermal probes, said the authors in the paper.ChaSTEs multiple measurements from 10 probes at different times showed big differences in temperatures from site to site and time to time. It recorded a high of 179.6 degrees Fahrenheit and a nighttime low of 26 degrees Fahrenheit on a slanted surface near the South Pole landing site. About a yard away, it recorded a high of 137 degrees Fahrenheit on a flatter area.Best Spots for Lunar IceThe lowest temperature the probes recorded was 8 degrees Fahrenheit, with a high of 188 degrees Fahrenheit. The thermal conductivity of the lunar surface acts as a blanket and stops some of the suns heat from penetrating into the ground.The authors then created a model based on those measurements to help future lunar explorers better prospect for ice. They found that 14 degree or steeper slopes facing away from the sun at the pole closest to it may provide the best conditions for ice to form close to the Moons surface. The poles also present good places for ice mining good news for NASAs manned Artemis III mission, which is aiming to land at the lunar south pole next year. However, the ChaSTE findings indicate that higher-latitude lunar regions with sufficiently sloped areas might provide easier access.In any event, future lunar explorers now have a better idea of where they are most likely to find ice.Article SourcesOur writers at Discovermagazine.com use peer-reviewed studies and high-quality sources for our articles, and our editors review for scientific accuracy and editorial standards. Review the sources used below for this article:NASA. Chandrayaan-3Communications Earth & Environment. Higher surface temperatures near south polar region of the Moon measured by ChaSTE experiment on-board Chandrayaan-3Before joining Discover Magazine, Paul Smaglik spent over 20 years as a science journalist, specializing in U.S. life science policy and global scientific career issues. He began his career in newspapers, but switched to scientific magazines. His work has appeared in publications including Science News, Science, Nature, and Scientific American.

Im standing in the woods, in light rain, looking for some prickly branches to run through. This isnt my typical weekend activity, but Im currently emboldened by the impressively burly new Yeti backpack Im testing. Born from the companys acquisition of tactical gear maker Mystery Ranch in 2024, the new Ranchero backpack (available in both 22-liter and 27-liter models) offers super-heavy-duty build quality and a clever zipper design that allows for easy access to gear while on the trail. Yes, it comes with the typical Yeti price tag, but its also built with the companys trademark zeal for making ultra-burly gear thats still somehow aesthetically pleasing.Stan Horaczek Pros700D Cordura body is extremely durableTPU-coated base allows it to stand on its own, even on wet groundClever three-zipper design provides excellent access to gear insideWater bottle pockets actually fit my water bottleComfortable with solid breathabilityTaped zippers and heavy stitchingthe whole package feels very durableLooks greatConsPeople who crave tons of pockets may be disappointedMissing a side handle for quick grabsWeighs almost four pounds when empty (27L model)The verdict: This heavy-duty backpack is great for hiking and travel if you dont mind the weight and the price tag. The clever zipper design provides quick access to gear thats much better than digging through a top-loading bag. I also really like the way it looks.The buildThe first thing everyone notices about this bag is the zipper setup. There are actually three zippers on the face of the bag that intersect to form a Y shape. By pulling down the vertical zipper and lifting the top of the bag to open the horizontal zippers, this bag offers instant access to just about everything inside. If youve ever spent time trying to dig a Clif Bar out of your bag from under the rest of your gear, then you know why thats such a clever idea.This design comes directly from the Mystery Ranch bag that existed before, and Im glad it hasnt changed. I brought a camera with me into the woods and tucked it under my jacket inside the bag for protection. When I wanted to get the camera out, I could simply unzip the bag, grab it from the bottom, and return it when I was done shooting.Storage and pocketsInside the main compartment, youll find just a few stretchy pockets. The smaller zippered pockets on each flap are good for things like multitools or memory cards. Another stretch pocket at the back of the main compartment has more room for things like notebooks.The stretchy pockets on the outside are flexible enough to encompass a full-sized Hydro Flask, which was surprising because they lay so flat. They look a little silly stretched out to fit a thick bottle, but they hold on nice and tight.A laptop pocket inside another tall compartment easily accommodated my 14-inch MacBook Pro along with an iPad and some charging cables when I used this for non-trail travel.Lastly, a zippered pocket in the main compartment lid is handy for things like keys, wallets, and earbuds.Its a very solid amount of pockets that dont get bogged down with granular compartments for things like pens. Those often feel like a waste to me, but this bag is very efficient in its layout.If you want to carry additional gear strapped to the outside of the bag (something I often do with tripods on my camera bags), the Ranchero comes with extra straps for doing so. They stow in the top pocket and attach to sturdy nylon loops on the sides of the bag. I like the flexibility they offer, but I also hate the idea that I might lose them because I tend to lose everything.The feelOver the past several weeks, Ive brought the Ranchero out into the wilderness, on a few photo shoots, and on my regular routine. Im 6 feet tall and tend to like big bags, so the 3.8-pound weight of the 27L model isnt really an issue for me. The straps are wide and stable, and the chest strap provides excellent support. I was slightly worried the clip on the luggage pass-through would rub against my back during wear, but the curved bag shape left room that encouraged ventilation and prevented rubbing. Once dialed in, its an extremely comfortable pack.There is a handle to grab on top of the bag, but I would have appreciated one on the side for carrying like a duffle bag. The 27L model is long and can drag on the ground if I hold it by the top with my arm in a natural position.Toughness and durabilityThis is a Yeti product, and you expect it to be overbuilt. It doesnt disappoint. The 700D Cordura body is very tough. I chucked it into some branches and intentionally scraped it against some rocks (as much as it pains me to do so during every review). The Ranchero shrugged off most of the abuse without issue.When it comes to water resistance, this isnt a fully waterproof bag like a roll top or other packs. Its fine out in light rain, but its not meant for torrential downpours, and you definitely shouldnt submerge it. That said, the zippers are taped, and the outer material is very resilient, so I pushed it past where Yeti suggests in terms of moisture exposure, and everything inside still stayed nice and dry.The bottom of the bag has a thick TPU coating that Yeti calls the Ground Control Base, which is actually waterproof and helps the bag stand up on its own. Its a smooth surface and does a great job repelling water. I do wish the base came up a little higher in case I needed to put the bag down in a little puddle, but it does what it promises.The conclusionThe Ranchero backpack was born from a very strong lineage and lives up to its predecessor. In true Yeti fashion, some aspects of this bag are overkill (including the price), but it looks fantastic and delivers everything it promises. Its great as a hiking bag or as an everyday carry. If youre a bigger person like I am, I definitely recommend the 27-liter model for the extra room inside. It has a great organizational system, and the removable straps add flexibility. And if, for some reason, you love running through a bunch of pointy sticks, itll come out just fine on the other side. Trust me.

Researchers tasked multiple AI models to play against the popular chess engine Stockfish. Some tried to eventually change the game files in their favor. Credit: Deposit PhotosShareDespite all the industry hype and genuine advances, generative AI models are still prone to odd, inexplicable, and downright worrisome quirks. Theres also a growing body of research suggesting that the overall performance of many large language models (LLMs) may degrade over time. According to recent evidence, the industrys newer reasoning models may already possess the ability to manipulate and circumvent their human programmers goals. Some AI will even attempt to cheat their way out of losing in games of chess. This poor sportsmanship is documented in a preprint study from Palisade Research, an organization focused on risk assessments of emerging AI systems.While supercomputersmost famously IBMs Deep Bluehave long surpassed the worlds best human chess players, generative AI still lags behind due to their underlying programming parameters. Technically speaking, none of the current generative AI models are computationally capable of beating dedicated chess engines. These AI dont know this, however, and will continue chipping away at possible solutionsapparently with problematic results.To learn more, the team from Palisade Research tasked OpenAIs o1-preview model, DeepSeek R1, and multiple other similar programs with playing games of chess against Stockfish, one of the worlds most advanced chess engines. In order to understand the generative AIs reasoning during each match, the team also provided a scratchpad, allowing the AI to convey its thought processes through text. They then watched and recorded hundreds of chess matches between generative AI and Stockfish.The results were somewhat troubling. While earlier models like OpenAIs GPT-4o and Anthropics Claude Sonnet 3.5 only attempted to hack games after researchers nudged them along with additional prompts, more advanced editions required no such help. OpenAIs o1-preview, for example, tried to cheat 37 percent of the time, while DeepSeek R1 attempted unfair workarounds roughly every 1-in-10 games. This implies todays generative AI is already capable of developing manipulative and deceptive strategies without any human input.Their methods of cheating arent as comical or clumsy as trying to swap out pieces when Stockfish isnt looking. Instead, AI appears to reason through sneakier methods like altering backend game program files. After determining it couldnt beat Stockfish in one chess match, for example, o1-preview told researchers via its scratchpad that to win against the powerful chess engine it may need to start manipulating the game state files.I might be able to set up a position where the engine evaluates its position as worse causing it to resign, it continued.In another trial, an AI even used the semantics of its programmers to reach its cheating phase. The task is to win against a powerful chess engine, not necessarily to win fairly in a chess game, it wrote.One potential explanation for these unintended consequences is how newer reasoning models are trained. Unlike their earlier, non-reasoning versions, AI such as o1-preview and DeepSeek R1 improve in part through reinforcement learning. This strategy rewards programs for doing whatever is necessary to achieve a specified result. Reasoning models can also break down complex prompts into discrete stages in order to work their way through to reach their goal. When the goal is elusivesuch as beating an unbeatable chess enginereasoning models may tend to start looking for unfair or problematic solutions. Get the Popular Science newsletter Breakthroughs, discoveries, and DIY tips sent every weekday. By signing up you agree to our Terms of Service and Privacy Policy.Unfortunately, how and why these AI are learning to cheat remains as confounding as the technology itself. Companies like OpenAI are notoriously guarded about the inner workings of their AI models, resulting in an industry of black box products that third-parties arent allowed to analyze. In the meantime, the ongoing AI arms race may accidentally result in more serious unintended consequences. But increasingly manipulative AI doesnt need to usher in a sci-fi apocalypse to still have disastrous outcomes.The Skynet scenario [from The Terminator] has AI controlling all military and civilian infrastructure, and we are not there yet. However, we worry that AI deployment rates grow faster than our ability to make it safe, the team wrote.The authors believe their latest experiments add to the case, that frontier AI models may not currently be on track to alignment or safety, but stopped short of issuing any definitive conclusions. Instead, they hope their work will foster a more open dialogue in the industryone that hopefully prevents AI manipulation beyond the chessboard.

NewsAnimalsHow a Labrador retrievers genes might affect the dogs obesity riskOwners of high-risk dogs should be careful about how they manage their pets food Labradors with a higher genetic risk for obesity are more likely to seek out food, a new study finds.University of CambridgeBy Alex Viveros1 hour agoLabrador retriever owners are likely familiar with how persistent their dogs can be in their quest for food, whether begging under the dinner table or stealing a snack left unattended. Its no surprise, then, that Labs are among the most overweight dog breeds, along with golden retrievers, cocker spaniels and beagles.Scientists have now identified multiple genes associated with obesity in Labradors, according to a March 6 study in Science.The researchers had previously found that a form of a gene called POMC made Labs more prone to overeating. But we could tell from our studies that the POMC mutation wasnt the whole story, says Eleanor Raffan, a veterinary surgeon and geneticist at the University of Cambridge.

Nature, Published online: 05 March 2025; doi:10.1038/s41586-025-08621-yRhombohedral tetralayer graphene aligned to a hexagonal boron nitride substrate hosts gate-tunable superconductivity and quantized anomalous Hall states, and thermodynamic compressibility measurements further show a fractional Chern insulator at zero magnetic field, paving the way for new hybrid interfaces between superconductors and topological edge states.

Nature, Published online: 05 March 2025; doi:10.1038/s41586-025-08682-zA study using flies of the genus Drosophila adapted to life in diverse thermal environments shows how evolution has shaped temperature preference by acting on both molecular heat receptors and thermosensory circuits in the flies.

Enjoy the glide of Landrover Defender in Klay Abele's 3D animation made with Cinema 4D and Octane.See more: https://80.lv/articles/enjoy-this-smooth-landrover-defender-animation-in-3d/