Meta Platforms, the Facebook and Instagram owner, faces pressure in Europe to ask before using data to target ads or train artificial intelligence.

He argued that offshoring allowed the U.S. semiconductor industry to create more high-paying jobs at home.

The actor stars in Peacocks gripping 10-part adaptation of the 1971 bestseller about an assassin, who in this telling is hired to kill a tech billionaire.

Karla Sofia Gascn, Zoe Saldaa and Selena Gomez star in Jacques Audiards operatic film about a drug-cartel kingpin who transitions from male to female.

An elephant never forgets This elephant figured out how to use a hose to shower A younger rival may have learned how to sabotage those showers by disrupting water flow. Jennifer Ouellette Nov 12, 2024 6:06 pm | 15 An elephant named Mary has been filmed using a hose to shower herself. Credit: Urban et al./Current Biology An elephant named Mary has been filmed using a hose to shower herself. Credit: Urban et al./Current Biology Story textSizeSmallStandardLargeWidth *StandardWideLinksStandardOrange* Subscribers only Learn more Mary the elephant shows off her hose-showering skills. Credit: Urban et al./Current Biology An Asian elephant named Mary living at the Berlin Zoo surprised researchers by figuring out how to use a hose to take her morning showers, according to a new paper published in the journal Current Biology. Elephants are amazing with hoses, said co-author Michael Brecht of the Humboldt University of Berlin. As it is often the case with elephants, hose tool use behaviors come out very differently from animal to animal; elephant Mary is the queen of showering.Tool use was once thought to be one of the defining features of humans, but examples of it were eventually observed in primates and other mammals. Dolphins have been observed using sea sponges to protect their beaks while foraging for food, and sea otters will break open shellfish like abalone with rocks. Several species of fish also use tools to hunt and crack open shellfish, as well as to clear a spot for nesting. And the coconut octopus collects coconut shells, stacking them and transporting them before reassembling them as shelter.Birds have also been observed using tools in the wild, although this behavior was limited to corvids (crows, ravens, and jays), although woodpecker finches have been known to insert twigs into trees to impale passing larvae for food. Parrots, by contrast, have mostly been noted for their linguistic skills, and there has only been limited evidence that they use anything resembling a tool in the wild. Primarily, they seem to use external objects to position nuts while feeding.And then there's Figaro, a precociousmale Goffin's cockatoo kept in captivity and cared for by scientists in the "Goffin lab" at the University of Veterinary Medicine in Vienna. Figaro showed a surprising ability to manipulate single tools to maneuver a tasty nut out of a box. Other cockatoos who repeatedly watched Figaro's performance were also able to do so. Figaro and his cockatoo cronies even learned how to combine toolsa stick and a ballto play a rudimentary form of "golf."Shower timeBoth captive and wild elephants are known to use and modify branches for fly switching. Brecht's Humboldt colleague, Lina Kaufman, is the one who first observed Mary using a hose to shower at the Berlin Zoo and told Brecht about it. They proceeded to undertake a more formal study of the behavior not just of Mary, but two other elephants at the zoo, Pang Pha and her daughter Anchali. Mary was born in the wild in Vietnam, while Pang Pha was a gift from Thailand; Anchali was born at the Berlin Zoo, where she was hand-raised by zookeepers. Showering was part of the elephants' morning routine, and all had been trained not to step on the hoses. Mary's rival Anchali blocking the flow of water. Urban et al./Current Biology All the elephants used their trunks to spray themselves with water, but Mary was the only one who also used the hose, picking it up with her trunk. Her hose showers lasted about seven minutes, and she dropped the hose when the water was turned off. Where she gripped the hose depended on which body part she was showering: she grasped it further from the end when spraying her back than when showering the left side of her body, for instance. This is a form of tool modification that has also been observed in New Caledonian crows.And the hose-showering behavior was "lateralized," that is, Mary preferred targeting her left body side more than her right. (Yes, Mary is a "left-trunker.") Mary even adapted her showering behavior depending on the diameter of the hose: she preferred showering with a 24-mm hose over a 13-mm hose and preferred to use her trunk to shower rather than a 32-mm hose.It's not known where Mary learned to use a hose, but the authors suggest that elephants might have an intuitive understanding of how hoses work because of the similarity to their trunks. "Bathing and spraying themselves with water, mud, or dust are very common behaviors in elephants and important for body temperature regulation as well as skin care," they wrote. "Mary's behavior fits with other instances of tool use in elephants related to body care."Perhaps even more intriguing was Anchali's behavior. While Anchali did not use the hose to shower, she nonetheless exhibited complex behavior in manipulating the hose: lifting it, kinking the hose, regrasping the kink, and compressing the kink. The latter, in particular, often resulted in reduced water flow while Mary was showering. Anchali eventually figured out how to further disrupt the water flow by placing her trunk on the hose and lowering her body onto it. Control experiments were inconclusive about whether Anchali was deliberately sabotaging Mary's shower; the two elephants had been at odds and behaved aggressively toward each other at shower times. But similar cognitively complex behavior has been observed in elephants.When Anchali came up with a second behavior that disrupted water flow to Mary, I became pretty convinced that she is trying to sabotage Mary, Brecht said. Do elephants play tricks on each other in the wild? When I saw Anchali's kink and clamp for the first time, I broke out in laughter. So, I wonder, does Anchali also think this is funny, or is she just being mean?Current Biology, 2024. DOI: 10.1016/j.cub.2024.10.017 (About DOIs).Jennifer OuelletteSenior WriterJennifer OuelletteSenior Writer Jennifer is a senior reporter at Ars Technica with a particular focus on where science meets culture, covering everything from physics and related interdisciplinary topics to her favorite films and TV series. Jennifer lives in Baltimore with her spouse, physicist Sean M. Carroll, and their two cats, Ariel and Caliban. 15 Comments Prev story

secret math problems dept. New secret math benchmark stumps AI models and PhDs alike FrontierMath's difficult questions remain unpublished so that AI companies can't train against it. Benj Edwards Nov 12, 2024 5:49 pm | 31 Credit: Getty Images Credit: Getty Images Story textSizeSmallStandardLargeWidth *StandardWideLinksStandardOrange* Subscribers only Learn moreOn Friday, research organization Epoch AI released FrontierMath, a new mathematics benchmark that has been turning heads in the AI world because it contains hundreds of expert-level problems that leading AI models solve less than 2 percent of the time, according to Epoch AI. The benchmark tests AI language models (such as GPT-4o, which powers ChatGPT) against original mathematics problems that typically require hours or days for specialist mathematicians to complete.FrontierMath's performance results, revealed in a preprint research paper, paint a stark picture of current AI model limitations. Even with access to Python environments for testing and verification, top models like Claude 3.5 Sonnet, GPT-4o, o1-preview, and Gemini 1.5 Pro scored extremely poorly. This contrasts with their high performance on simpler math benchmarksmany models now score above 90 percent on tests like GSM8K and MATH.The design of FrontierMath differs from many existing AI benchmarks because the problem set remains private and unpublished to prevent data contamination. Many existing AI models are trained on other test problem datasets, allowing the AI models to easily solve the problems and appear more generally capable than they actually are. Many experts cite this as evidence that current large language models (LLMs) are poor generalist learners.Problems spanning multiple disciplinesEpoch AI says it developed FrontierMath through collaboration with over 60 mathematicians from leading institutions. The problems underwent peer review to verify correctness and check for ambiguities. About 1 in 20 problems needed corrections during the review process, a rate comparable to other major machine learning benchmarks.The problems in the new set span multiple mathematical disciplines, from computational number theory to abstract algebraic geometry. And they are reportedly difficult to solve. Really, really difficult.Epoch AI allowed Fields Medal winners Terence Tao and Timothy Gowers to review portions of the benchmark. "These are extremely challenging," Tao said in feedback provided to Epoch. "I think that in the near term basically the only way to solve them, short of having a real domain expert in the area, is by a combination of a semi-expert like a graduate student in a related field, maybe paired with some combination of a modern AI and lots of other algebra packages." A chart showing AI models' limited success on the FrontierMath problems, taken from Epoch AI's research paper. Credit: Epoch AI To aid in the verification of correct answers during testing, the FrontierMath problems must have answers that can be automatically checked through computation, either as exact integers or mathematical objects. The designers made problems "guessproof" by requiring large numerical answers or complex mathematical solutions, with less than a 1 percent chance of correct random guesses.Mathematician Evan Chen, writing on his blog, explained how he thinks that FrontierMath differs from traditional math competitions like the International Mathematical Olympiad (IMO). Problems in that competition typically require creative insight while avoiding complex implementation and specialized knowledge, he says. But for FrontierMath, "they keep the first requirement, but outright invert the second and third requirement," Chen wrote.While IMO problems avoid specialized knowledge and complex calculations, FrontierMath embraces them. "Because an AI system has vastly greater computational power, it's actually possible to design problems with easily verifiable solutions using the same idea that IOI or Project Euler doesbasically, 'write a proof' is replaced by 'implement an algorithm in code,'" Chen explained.The organization plans regular evaluations of AI models against the benchmark while expanding its problem set. They say they will release additional sample problems in the coming months to help the research community test their systems.Benj EdwardsSenior AI ReporterBenj EdwardsSenior AI Reporter Benj Edwards is Ars Technica's Senior AI Reporter and founder of the site's dedicated AI beat in 2022. He's also a widely-cited tech historian. In his free time, he writes and records music, collects vintage computers, and enjoys nature. He lives in Raleigh, NC. 31 Comments

The company -- taking direct aim at OpenAI, Anthropic, and other incumbents in the GenAI arms race -- plans to use the funding to fuel its agentic AI efforts.

Lisa Morgan, Freelance WriterNovember 12, 20246 Min ReadDragos Condrea via Alamy StockSoftware development is an ever-changing landscape. Over the years, it has become easier to generate high-quality code faster, though the definition of faster is a moving target.Take low-code tools, for example. With them, developers can build most of the functionality they need with the platform, so they only need to write the custom code the application requires. Low-code tools have also democratized software development -- particularly with the addition of AI.GenAI is accelerating development even further, and its changing the way developers think about code.Siddharth Parakh, senior engineering manager at Medable, expects Ai to revolutionize productivity.The ability for AI to automate repetitive tasks, refactor code and even generate solutions from scratch would allow developers to focus on higher-order problem-solving and strategic design decisions, says Parakh in an email interview. With AI handling routine coding, developers could become orchestrators of complex systems rather than line-by-line authors of software.But theres a catch: Currently, AI-generated code cannot fully replace human intuition in areas such as creative problem solving, contextual understanding, and domain-specific decision-making. Also, AI models are only as good as the data they are trained on, which can lead to bias issues, error propagation or unsafe coding practices, he says. Quality control, debugging, and nuanced decision-making are still areas where human expertise is necessary.Related:How AI HelpsThe operative work is automation.If AI takes over the majority of coding tasks, it would drive unprecedented efficiency and speed in software development, says Medables Parakh. Teams could iterate faster, adapt to changes more fluidly and scale projects without the traditional bottlenecks of manual coding. This could democratize software development, enabling non-experts to create functional software with minimal input.Geoffrey Bourne, co-founder of social media API company Ayrshare, says GenAI coding assistants are now an integral part of his coding.They produce lines of code which save me hours on a weekly basis. But, although the results are improving, theyre correct less than 40% of the time. You need the experience to know the code just isnt up to scratch and needs adjusting or a redo, says Bourne in an email interview. Newbie coders are starting out with these assistants at their fingertips but without the years of experience writing code their seniors have. Weve got to take this into account and not necessarily limit their access but find creative ways to inject that knowledge. You need to find a balance [between] the instant code fix with healthy experience and a critical eye.Related:The evolution of programming, especially through abstraction layers and GenAI, has significantly transformed the way Surabhi Bhargava, a machine learning tech leadat Adobe, approaches her work.GenAI has made certain aspects of development much faster. Writing boilerplate code, prototyping and even debugging is now more streamlined. Finding information across different documents is easier with AI and copilots, says Bhargava in an email interview. [Though] AI can speed things up, I now [must] critically assess AI-generated outputs. It has made me more analytical in reviewing the work produced by these systems, ensuring it aligns with my expectations and needs, particularly when handling complex algorithms or compliance-driven work.AI tools are also helping her create rapid prototypes and theyre reducing the cognitive load.I can focus more on strategic thinking, which improves productivity and gives me room to innovate, says Bhargava. Sometimes, its tempting to lean too heavily on AI for code generation or decision-making. AI-generated solutions arent always optimized or tailored for the specific needs of a project, resulting in bugs and issues in prod. [And] sometimes, it takes more time to set it up if the tools are complex to use.Related:Hands-Free Coding Still Hasnt ArrivedAt present, AI struggles with its own set of issues such as misinterpretation, hallucination and incorrect facts. Over-reliance on AI-generated code could lead to a lack of deep technical expertise in development teams.With humans less involved in the nitty-gritty of coding, we could see a decline in the essential skills needed to debug, optimize, or creatively problem-solve at a low level. Additionally, ethical and security concerns could arise as AI systems might unknowingly introduce vulnerabilities or generate biased solutions, says Parakh. Tom Taulli, author of AI-Assisted Programming: Better Planning, Coding, Testing, and Deployment has been using AI-assisted programming tools for the past couple years. This technology has had the most transformative impact by far on his work in his over 40-year work history.Whats interesting is that I approach a project in terms of natural language prompts, not coding or doing endless searchers on Google and StackOverflow. In fact, I set up a product requirements document that is a list of prompts. Then, I go through each one for the development of an application, says Taulli. These systems are far from perfect. But it only takes a few seconds to generate the code -- and this means I have more time to review it and make iterations.Taulli has been a backend developer primarily, but AI assisted programming has allowed him to do more front-end development.The funny thing is that one of the biggest drawbacks is the pace of innovation with these tools. It can be tough to keep up with the many developments, says Taulli. True, there are other well-known disadvantages, such as with security and intellectual property. Is the code being copied? Do you really own the code you create? says Taulli. However, I think one of the biggest drawbacks is the context window. Basically, the LLMs cannot understand the large codebases. This can make it difficult for sophisticated code refactoring..Another issue is the cut-off date of the LLMs. They may not have the latest packages and frameworks, but the benefits outweigh the drawbacks, he says.Tom Jauncey, head nerd at digital marketing agency Nautilus Marketing, says GenAI tools like GitHub Copilot have accelerated the coding process by letting him think about high-level architecture and design. His advice is to use AI to save time on boilerplate code and documentation.Some of the things that I had to learn were how to prompt AI tools and think critically about their output. It is important to remember that while AI is great at generative code, it doesn't always understand broader context and business requirements, says Jauncey. Thus, always cross-check the AI-generated code with official documentation. AI-powered tools ease the effort of exploring a new language or framework without having to go into syntax details.Edward Tian, CEO of GPTZero, believes its better to use GenAI to assist coding rather than relying on it entirely.Personalization is such a key aspect of coding, and GenAI sometimes just cant quite personalize things in the way you want. It can certainly create complicated code, but it just often falls short in terms of uniqueness, says Tian.Bottom LineGenAI is accelerating development by generating code quickly but beware of its limitations. While its good for writing boilerplate code and documentation, creating quick prototypes and debugging, its important to verify the outputs. Prompt engineering skills also help boost productivity.About the AuthorLisa MorganFreelance WriterLisa Morgan is a freelance writer who covers business and IT strategy and emergingtechnology for InformationWeek. She has contributed articles, reports, and other types of content to many technology, business, and mainstream publications and sites including tech pubs, The Washington Post and The Economist Intelligence Unit. Frequent areas of coverage include AI, analytics, cloud, cybersecurity, mobility, software development, and emerging cultural issues affecting the C-suite.See more from Lisa MorganNever Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.SIGN-UPYou May Also LikeWebinarsMore WebinarsReportsMore Reports

Wildfires in the tropics drove some increase in CO2 emissions but the bulk was driven by burning fossil fuelsCarl De Souza/AFP/Getty ImagesCarbon dioxide emissions from burning fossil fuels in 2024 are set to blow past last years record levels, dashing hopes this year will see the planet-warming emissions peak.Reducing emissions is more urgent than ever and theres only one way to do it: massively reduce fossil emissions, says Pierre Friedlingstein at the University of Exeter, UK. AdvertisementThat is according to the latest Global Carbon Budget report, a preliminary accounting of CO2 emissions to date with projections to the end of the year, produced by Friedlingstein and his colleagues. It was released at the COP29 summit now underway in Azerbaijan, where countries aim to set new financial targets to address climate change.Last year, some researchers were forecasting a peak in emissions in 2024, but the report finds human-caused CO2 emissions are set to reach a record 41.6 gigatonnes in 2024, a 2 per cent rise on 2023s record. Almost 90 per cent of that total consists of emissions from burning fossil fuels. The rest is from changes in the land driven mostly by deforestation and wildfires.At 0.8 per cent, the growth rate of fossil fuel emissions is half that of 2023, although it remains higher than the average rate over the past decade. [The slower rate] is a good sign, but its still miles away from where we need to get, says Friedlingstein. Get a dose of climate optimism delivered straight to your inbox every month.Sign up to newsletterDespite a long-term downward trend, projected emissions from land use change also increased this year, largely due to drought-driven wildfires in the tropics. Some of the increase is also down to a collapse of the carbon land sink in 2023, which usually removes about a quarter of our annual CO2 emissions from the atmosphere. This sink declined by more than 40 per cent last year and the early part of 2024 as global temperatures spiked under the influence of El Nio.2023 is an incredible demonstration of what can happen in a warmer world when we had peak records in global temperatures combined with El Nio droughts and fires, says Pep Canadell at the Commonwealth Scientific and Industrial Research Organisation in Australia, a co-author of the report. Put all these things together and last year we had almost a third less help removing atmospheric CO2 by the worlds forests than we have had over the last decade.While this also added to emissions in 2024, the researchers expect this land carbon sink has mostly recovered as the warming influence of El Nio has faded. Its not a long-term collapse, says Friedlingstein.The report finds CO2 emissions in China, which generates nearly a third of the global total, are only projected to increase by 0.2 per cent in 2024 compared to 2023. Canadell says that because of the large margin of error in this projection of Chinas emissions, it is actually possible they have stayed steady or gone down. Indias emissions also increased at a slower rate than last year, rising by just under 5 per cent. In the US and the EU, emissions continued to decline, albeit at a much slower rate than last year.Hot temperatures that boost electricity demand to power air conditioning are also a key reason why fossil fuel emissions have continued to rise despite the massive build-out of renewables in 2024, says Neil Grant at Climate Analytics, a think tank in Germany. Whether due to electric vehicles, data centres or manufacturing, most people have been caught a bit surprised by the level of electricity demand this year, he says.If emissions continue at this level, the report finds that within six years the world will exceed its remaining carbon budget to limit warming to 1.5C above pre-industrial levels, and will exceed the budget to stay within 2C warming within 27 years.We have to accelerate, accelerate, accelerate, accelerate the transition to renewable energy, says Candell. Climate change is like a slippery slope that we can just keep falling down. We need to slam on the brakes as hard as we can so we can stop falling.Topics:

Eurasian reed warblers migrate between Europe and AfricaAGAMI Photo Agency / Alamy StockMany migratory birds use Earths magnetic field as a compass, but some can also use information from that field to determine more or less where they are on a mental map.Eurasian reed warblers (Acrocephalus scirpaceus) appear to calculate their geographical position by drawing data from different distances and angles between magnetic fields and the Earths shape. The findings suggest that the birds use magnetic information as a sort of GPS that tells them not only where to go, but where they are initially, says Richard Holland at Bangor University in the UK. AdvertisementWhen we travel, we have a map which tells us where we are and we have a compass, which tells us which way to go to reach our destination, he says. We dont think birds have quite this level of accuracy or degree of knowledge of the whole Earth. Even so, they see how magnetic cues change as they move along their normal path or even if theyre far displaced from that path.Scientists have known for decades that migratory birds rely on cues from the sun, the stars and Earths magnetic field to determine which direction to head towards. But figuring out direction using a compass is markedly different from knowing where in the world they are, and scientists still debate about whether and how birds figure out their current map position.Florian Packmor at Lower Saxon Wadden Sea National Park Authority in Germany suspected birds could detect detailed aspects of the magnetic field to determine their global position. Specifically, he thought they might use magnetic inclination the changing angle of Earths surface relative to its magnetic lines and magnetic declination the difference in direction between the geographic and magnetic poles to understand more precisely where they are located in the world. A monthly celebration of the biodiversity of our planets animals, plants and other organisms.Sign up to newsletterTo test that theory, Packmor, Holland and their colleagues captured 21 adult reed warblers on their migration route from Europe to Africa in Illmitz, Austria. There, they placed the birds temporarily in outdoor aviaries, where the researchers used a Helmholtz coil to interfere with magnetic fields. They artificially altered the inclination and declination in a way that corresponded to a position in Neftekamsk, Russia, 2600 kilometres away. Thats way out of their direction, says Packmor.The team then put the birds in a special cage for studying migratory instincts and asked two independent researchers who were unaware of the changes in magnetic field to record which way the birds headed. In the modified magnetic field situations, most of the birds showed a clear penchant for flying west-southwest, as though they were trying to return to their migration route from Russia. By contrast, the same birds wanted to fly south-southeast out of Austria when the magnetic field was unmodified.This suggests that the birds believed that they were no longer in Austria, but in Russia based on their magnetic inclination and declination alone, says Packmor.Of course, they dont know its Russia, but its too far north and east of where they should be, says Holland. And then at that point, they look at their compass system to work out how to fly south and west.However, we still dont fully understand the neurological mechanisms that enable birds to sense these aspects of Earths magnetic field.This is an important step in understanding how magnetic maps of songbirds and in particular, reed warblers work, says Nikita Chernetsov at the Zoological Institute of the Russian Academy of Sciences in Saint Petersburg, who was not involved in the study.While the research confirms reed warblers rely on these magnetic fields for positioning, it doesnt mean that all birds do so, he adds. Not all birds work the same way.The birds were released two to three weeks after the study, at which time they could continue their normal migration, Packmor and Holland say. Indeed, one of the birds they studied was captured a second time a year later, meaning the teams research did not prevent it from migrating successfully.Journal reference:Proceedings of the Royal Society B: Biological Sciences DOI: 10.1098/rspb.2024.1363Topics: