Linux fréttir

Analysts say production will top out this decade while global electrification keeps ramping

Concerns are mounting over copper supplies, with a fresh study warning that demand will likely outstrip production within a decade, threatening to constrain global technological advancement.…

The world's oceans absorbed yet another record-breaking amount of heat in 2025, continuing an almost unbroken streak of annual records since the start of the millennium and fueling increasingly extreme weather events around the globe. More than 90% of the heat trapped by humanity's carbon emissions ends up in the oceans, making ocean heat content one of the clearest indicators of the climate crisis's trajectory. The analysis, published in the journal Advances in Atmospheric Sciences, drew on temperature data collected across the oceans and collated by three independent research teams. The measurements cover the top 2,000 meters of ocean depth, where most heat absorption occurs. The amount of heat absorbed is equivalent to more than 200 times the total electricity used by humans worldwide. This extra thermal energy intensifies hurricanes and typhoons, produces heavier rainfall and greater flooding, and results in longer marine heatwaves that decimate ocean life. The oceans are likely at their hottest in at least 1,000 years and heating faster than at any point in the past 2,000 years.

Read more of this story at Slashdot.

Huntress analysis suggests VM escape bugs were already weaponized in the wild

Chinese-linked cybercriminals were sitting on a working VMware ESXi hypervisor escape kit more than a year before the bugs it relied on were made public.…

The software wasn't actually renamed, but you couldn't be blamed for being confused

Opinion Wait? What? I was just cruising along the information superhighway – yes, I'm old, deal with it – when I spotted a Y Combinator story announcing, "Microsoft Office renamed to 'Microsoft 365 Copilot app'." Excuse me!? I looked closer and found that, sure enough, it certainly looked like Microsoft had renamed Office to the God-awful "Microsoft 365 Copilot."…

Extremophile bacteria could help turn Martian dirt into building material for human habitats

Tough microbes able to survive extreme environments on Earth could be the key to constructing buildings to allow humans to survive on Mars, according to a research paper.…

The queue might move on, but the software never did

Bork!Bork!Bork! Today's bork - on a UK border control wait-time screen - is doubly unfortunate. Tired passengers get no clue how long until someone checks their passport, and of all organizations that should keep security certs current, the one responsible for keeping out criminals tops the list.…

Image generation paywalled on X after ministers and regulators start asking awkward questions

Grok has yanked its image-generation toy out of the hands of most X users after the UK government openly weighed a ban over the AI feature that "undressed" people on command.…

alternative_right shares a report from ScienceAlert: At the Experimental Advanced Superconducting Tokamak (EAST), physicists successfully exceeded what is known as the Greenwald limit, a practical density boundary beyond which plasmas tend to violently destabilize, often damaging reactor components. For a long time, the Greenwald limit was accepted as a given and incorporated into fusion reactor engineering. The new work shows that precise control over how the plasma is created and interacts with the reactor walls can push it beyond this limit into what physicists call a 'density-free' regime. [...] A team led by physicists Ping Zhu of Huazhong University of Science and Technology and Ning Yan of the Chinese Academy of Sciences designed an experiment to take this theory further, based on a simple premise: that the density limit is strongly influenced by the initial plasma-wall interactions as the reactor starts up. In their experiment, the researchers wanted to see if they could deliberately steer the outcome of this interaction. They carefully controlled the pressure of the fuel gas during tokamak startup and added a burst of heating called electron cyclotron resonance heating. These changes altered how the plasma interacts with the tokamak walls through a cooler plasma boundary, which dramatically reduced the degree to which wall impurities entered the plasma. Under this regime, the researchers were able to reach densities up to about 65 percent higher than the tokamak's Greenwald limit. This doesn't mean that magnetically confined plasmas can now operate with no density limits whatsoever. However, it does show that the Greenwald limit is not a fundamental barrier and that tweaking operational processes could lead to more effective fusion reactors. The findings have been published in Science Advances.

Read more of this story at Slashdot.

Initial £7M estimate proves optimistic after multiple contract uplifts

The Bank of England has trebled the amount it is spending on its Oracle systems integrator amid efforts to migrate business applications to the cloud.…

As you should, when being told the only remedy is deleting everything and starting again

On Call 2025 has ended and a new year is upon us, but The Register will continue opening Friday mornings with a fresh installment of On Call – the reader-contributed column that tells your tales of tech support.…

Researchers at Stanford University have created a programmable synthetic "skin" that can independently change color and texture, "a feat previously only available within the animal kingdom," reports the Register. From the report: The technique employs electron beams to write patterns and add optical layers that create color effects. When exposed to water, the film swells to reveal texture and colors independently, depending on which side of the material is exposed, according to a paper published in the scientific journal Nature this week. In an accompanying article, University of Stuttgart's Benjamin Renz and Na Liu said the researchers' "most striking achievement was a photonic skin in which color and texture could be independently controlled, mirroring the separate regulation... in octopuses." The research team used the polymer PEDOT:PSS, which can swell in water, as the basis for their material. Its reaction to water can be controlled by irradiating it with electrons, creating textures and patterns in the film. By adding thin layers of gold, the researchers turned surface texture into tunable optical effects. A single layer could be used to scatter light, giving the shiny metal a matte, textured appearance. To control color, a polymer film was sandwiched between two layers of gold, forming an optical cavity, which selectively reflects light.

Read more of this story at Slashdot.

Nobody really needs an AI toothbrush that sends their gums to the cloud

Opinion Another Consumer Electronics Show has rolled through Las Vegas, and this year vendors scrawled “AI-enabled” on all the kit they hope will find its way into your home – while airbrushing away its immaturity and downsides.…

Outages hit Russia and Ukraine, too

The authors of a hypothetical manual containing procedures repressive governments can use to stay in power despite restive populations would surely devote its first chapter to turning off the internet, an action the government of Iran appears to have taken in the last 24 hours.…

Grab some popcorn for the Xi vs Zuck bout, which may not be the biggest fight on the card

Chinese authorities have signalled they’ll likely probe Meta’s planned acquisition of made-in-China AI platform Manus.…

An anonymous reader quotes a report from NPR: [I]t turns out that some genius dogs can learn a brand new word, like the name of an unfamiliar toy, by just overhearing brief interactions between two people. What's more, these "gifted" dogs can learn the name of a new toy even if they first hear this word when the toy is out of sight -- as long as their favorite human is looking at the spot where the toy is hidden. That's according to a new study in the journal Science. "What we found in this study is that the dogs are using social communication. They're using these social cues to understand what the owners are talking about," says cognitive scientist Shany Dror of Eotvos Lorand University and the University of Veterinary Medicine, Vienna. "This tells us that the ability to use social information is actually something that humans probably had before they had language," she says, "and language was kind of hitchhiking on these social abilities." [...] "There's only a very small group of dogs that are able to learn this differentiation and then can learn that certain labels refer to specific objects," she says. "It's quite hard to train this and some dogs seem to just be able to do it." [...] To explore the various ways that these dogs are capable of learning new words, Dror and some colleagues conducted a study that involved two people interacting while their dog sat nearby and watched. One person would show the other a brand new toy and talk about it, with the toy's name embedded into sentences, such as "This is your armadillo. It has armadillo ears, little armadillo feet. It has a tail, like an armadillo tail." Even though none of this language was directed at the dogs, it turns out the super-learners registered the new toy's name and were later able to pick it out of a pile, at the owner's request. To do this, the dogs had to go into a separate room where the pile was located, so the humans couldn't give them any hints. Dror says that as she watched the dogs on camera from the other room, she was "honestly surprised" because they seemed to have so much confidence. "Sometimes they just immediately went to the new toy, knowing what they're supposed to do," she says. "Their performance was really, really high." She and her colleagues wondered if what mattered was the dog being able to see the toy while its name was said aloud, even if the words weren't explicitly directed at the dog. So they did another experiment that created a delay between the dog seeing a new toy and hearing its name. The dogs got to see the unfamiliar toy and then the owner dropped the toy in a bucket, so it was out of sight. Then the owner would talk to the dog, and mention the toy's name, while glancing down at the bucket. While this was more difficult for dogs, overall they still could use this information to learn the name of the toy and later retrieve it when asked. "This shows us how flexible they are able to learn," says Dror. "They can use different mechanisms and learn under different conditions."

Read more of this story at Slashdot.

YouTube is updating search filters so users can explicitly choose between Shorts and long-form videos. The change also replaces view-count sorting with a new "Popularity" filter and removes underperforming options like "Sort by Rating." The Verge reports: Right now, a filter-less search shows a mix of longform and short form videos, which can be annoying if you just want to see videos in one format or the other. But in the new search filters, among other options, you can pick to see "Videos," which in my testing has only showed a list of longform videos, or "Shorts," which just shows Shorts. YouTube is also removing the "Upload Date - Last Hour" and "Sort by Rating" filters because they "were not working as expected and had contributed to user complaints." The company will still offer other "Upload Date" filters, like "Today," "This week," "This Month," and "This Year," and you can also find popular videos with the new "Popularity" filter, which is replacing the "View count" sort option. (With the new "Popularity" filter, YouTube says that "our systems assess a video's view count and other relevance signals, such as watch time, to determine its popularity for that specific query.")

Read more of this story at Slashdot.

Longtime Slashdot reader schwit1 shares a report from Reuters: Billionaire entrepreneur Elon Musk persuaded a judge on Wednesday to allow a jury trial on his allegations that ChatGPT maker OpenAI violated its founding mission in its high-profile restructuring to a for-profit entity. Musk was a cofounder of OpenAI in 2015 but left in 2018 and now runs an AI company that competes with it. U.S. District Judge Yvonne Gonzalez Rogers in Oakland, California, said at a hearing that there was "plenty of evidence" suggesting OpenAI's leaders made assurances that its original nonprofit structure was going to be maintained. The judge said there were enough disputed facts to let a jury consider the claims at a trial scheduled for March, rather than decide the issues herself. She said she would issue a written order after the hearing that addresses OpenAI's bid to throw out the case. [...] Musk contends he contributed about $38 million, roughly 60% of OpenAI's early funding, along with strategic guidance and credibility, based on assurances that the organization would remain a nonprofit dedicated to the public benefit. The lawsuit accuses OpenAI co-founders Sam Altman and Greg Brockman of plotting a for-profit switch to enrich themselves, culminating in multibillion-dollar deals with Microsoft and a recent restructuring. OpenAI, Altman and Brockman have denied the claims, and they called Musk "a frustrated commercial competitor seeking to slow down a mission-driven market leader." Microsoft is also a defendant and has urged the judge to toss Musk's lawsuit. A lawyer for Microsoft said there was no evidence that the company "aided and abetted" OpenAI. OpenAI in a statement after the hearing said: "Mr Musk's lawsuit continues to be baseless and a part of his ongoing pattern of harassment, and we look forward to demonstrating this at trial."

Read more of this story at Slashdot.

Dark copyright evasion magic makes light work of developers' guardrails

Machine learning models, particularly commercial ones, generally do not list the data developers used to train them. Yet what models contain and whether that material can be elicited with a particular prompt remain matters of financial and legal consequence, not to mention ethics and privacy.…

Resets 'Days Since Last Outage' counter to zero, aspires to do the same for everyone, forever

Analytics outfit Snowflake is buying telemetry data platform Observe to help its customers discover and mitigate IT issues before they cause downtime. It announced the deal on the same day its own services experienced a “major outage.”…

Illinois Department of Human Services disclosed that a misconfigured internal mapping website exposed sensitive personal data for more than 700,000 Illinois residents for over four years, from April 2021 to September 2025. Officials say they can't confirm whether the publicly accessible data was ever viewed. TechCrunch reports: Officials said the exposed data included personal information on 672,616 individuals who are Medicaid and Medicare Savings Program recipients. The data included their addresses, case numbers, and demographic data -- but not individuals' names. The exposed data also included names, addresses, case statuses, and other information relating to 32,401 individuals in receipt of services from the department's Division of Rehabilitation Services.

Read more of this story at Slashdot.