news aggregator

Simon Willison, commenting on the recent paper from Apple researchers that found state-of-the-art large language models face complete performance collapse beyond certain complexity thresholds: I thought this paper got way more attention than it warranted -- the title "The Illusion of Thinking" captured the attention of the "LLMs are over-hyped junk" crowd. I saw enough well-reasoned rebuttals that I didn't feel it worth digging into. And now, notable LLM skeptic Gary Marcus has saved me some time by aggregating the best of those rebuttals together in one place! [...] And therein lies my disagreement. I'm not interested in whether or not LLMs are the "road to AGI". I continue to care only about whether they have useful applications today, once you've understood their limitations. Reasoning LLMs are a relatively new and interesting twist on the genre. They are demonstrably able to solve a whole bunch of problems that previous LLMs were unable to handle, hence why we've seen a rush of new models from OpenAI and Anthropic and Gemini and DeepSeek and Qwen and Mistral. They get even more interesting when you combine them with tools. They're already useful to me today, whether or not they can reliably solve the Tower of Hanoi or River Crossing puzzles.

Read more of this story at Slashdot.

Case over bundling Play Store with Chrome and Google Search set to continue

A European court has advised [PDF] that Google's appeal against a ruling that found it had abused its market dominance should be dismissed.…

Apple software chief Craig Federighi has ruled out bringing macOS to the iPad, amusingly using a kitchen utensil analogy to explain the company's design philosophy. "We don't want to create a boat car or, you know, a spork," Federighi said in an interview. "Someone said, 'If a spoon's great, a fork's great, then let's combine them into a single utensil, right?' It turns out it's not a good spoon and it's not a good fork. It's a bad idea. And so we don't want to build sporks." The new version of iPadOS, which will ship to consumers later this year, features dynamically resizable windows that users can drag by their corners and a menu bar that is accessible through swipe gestures or cursor movement. Some observers might consider the iPad Pro itself a "convertible" product that blurs the line between tablet and laptop, he said. However, the Mac and iPad serve distinct purposes, he asserted. "The Mac lets the iPad be iPad," he said adding that Apple's objective "has not been to have iPad completely displace those places where the Mac is the right tool for the job." Rather than full convergence, Federighi said the iPad "can be inspired by elements of the Mac" while remaining a separate platform. "I think the Mac can be inspired by elements of iPad, and I think that that's happened a great deal."

Read more of this story at Slashdot.

Feds told they can't demand a haystack to find a needle

The United States is requesting [PDF] a month-long extension to the deadline for its final decision regarding an appeal against a judge's ruling that obtaining tower dumps is unconstitutional.…

Margaret Mitchell, chief ethics scientist at Hugging Face and founder of Google's responsible AI team, has dismissed artificial general intelligence as "just vibes and snake oil." Mitchell, who was ousted from Google in 2021, has co-written a paper arguing that AGI should not serve as a guiding principle for the AI industry. Mitchell contends that both "intelligence" and "general" lack clear definitions in AI contexts, creating what she calls an "illusion of consensus" that allows technologists to pursue any development path under the guise of progress toward AGI. "But as for now, it's just like vibes, vibes and snake oil, which can get you so far. The placebo effect works relatively well," she told FT in an interview. She warns that current AI advancement is creating a "massive rift" between those profiting from the technology and workers losing income as their creative output gets incorporated into AI training data.

Read more of this story at Slashdot.

Iranian state television has instructed residents to delete WhatsApp from their smartphones, claiming the messaging platform gathers user information to share with Israel. The local media provided no evidence supporting these allegations but additionally encouraged residents to avoid other "location-based" apps. WhatsApp has disputed the claims, with a spokesperson telling Time magazine the Meta-owned platform uses end-to-end encryption and does not track precise locations, keep messaging logs, or provide bulk information to governments. The episode comes at a time when Iran is simultaneously experiencing a "near-total national Internet blackout," according to NetBlock, an internet governance monitoring organization. The disruption follows earlier partial outages amid escalating military tensions with Israel after days of missile strikes between the countries. Further reading, from earlier this week: Iran Bans Officials From Using Internet-Connected Devices.

Read more of this story at Slashdot.

Observatory maps datacenter hubs and power demand worldwide

The International Energy Agency (IEA) has unveiled an online platform to closely monitor and analyze the impact of AI across the energy sector worldwide.…

93 Escort Wagon writes: More proof that Amazon's leadership views the balance of power between itself and its workforce tilting decisively in its favor: Amazon's employees are being told they must relocate to one of the company's large hubs -- with the company specifying the required location -- or resign with no severance. CEO Andy Jassy did have the grace to give people 30 days to decide.

Read more of this story at Slashdot.

ChatGPT-assisted writing dampened brain activity and recall in a controlled MIT study [PDF] of 54 college volunteers divided into AI-only, search-engine, and no-tool groups. Electroencephalography recorded during three essay-writing sessions found the AI group consistently showed the weakest neural connectivity across all measured frequency bands; the tool-free group showed the strongest, with search users in between. In the first session 83% of ChatGPT users could not quote any line they had just written and none produced a correct quote. Only nine of the 18 claimed full authorship of their work, compared with 16 of 18 in the brain-only cohort. Neural coupling in the AI group declined further over repeated use. When these participants were later asked to write without assistance, frontal-parietal networks remained subdued and 78% again failed to recall a single sentence accurately. The pattern reversed for students who first wrote unaided: introducing ChatGPT in a crossover session produced the highest connectivity sums in alpha, theta, beta and delta bands, indicating intense integration of AI suggestions with prior knowledge. The MIT authors warn that habitual reliance on large language models "accumulates cognitive debt," trading immediate fluency for weaker memory, reduced self-monitoring, and narrowed neural engagement.

Read more of this story at Slashdot.

Experts note 'major red flags' in donut giant's security as 161,676 staff and families informed of attack details

Krispy Kreme finally revealed the number of people affected by its November cyberattack, and it's easy to see why analyzing the incident took the well-resourced company several months.…

An anonymous reader quotes a report from The Guardian: Foreign students will be required to unlock their social media profiles to allow US diplomats to review their online activity before receiving educational and exchange visas, the state department has announced. Those who fail to do so will be suspected of hiding that activity from US officials. The new guidance, unveiled by the state department on Wednesday, directs US diplomats to conduct an online presence review to look for "any indications of hostility toward the citizens, culture, government, institutions, or founding principles of the United States." A cable separately obtained by Politico also instructs diplomats to flag any "advocacy for, aid or support for foreign terrorists and other threats to US national security" and "support for unlawful antisemitic harassment or violence." The screening for "antisemitic" activity matches similar guidance given at US Citizenship and Immigration Services under the Department of Homeland Security and has been criticized as an effort to crack down on opposition to the conduct of Israel's war in Gaza. The new state department checks are directed at students and other applicants for visas in the F, M and J categories, which refer to academic and vocational education, as well as cultural exchanges. "It is an expectation from American citizens that their government will make every effort to make our country safer, and that is exactly what the Trump administration is doing every single day," said a senior state department official, adding that Marco Rubio was "helping to make America and its universities safer while bringing the state Department into the 21st century."

Read more of this story at Slashdot.

Veteran X-ray telescope discovery shows that... phew, current model of the cosmos still works

Astronomers have found a filament of hot gas, ten times as massive as our galaxy, that they reckon could explain where at least some of the universe's "missing" matter might be lurking.…

Good to see government that values its academics (cough cough). Plus: New board criticized for lacking 'ops' people

Cybersecurity experts have started a formal review into the UK cybersecurity market, at the government's request, to identify future growth opportunities as it looks to grow the industry that's core to the country's Industrial Strategy.…

Test fire trouble means Musk's rocketeers reset the 'days since Starship had a major anomaly' counter to zero

SpaceX has made excellent progress with its Starship rocket. The stainless steel vehicle can now explode before even leaving the Earth.…

Valve has quietly updated the Steam Beta Client to enable Proton by default for all Windows games on Linux, eliminating the need for users to toggle compatibility settings manually. GamingOnLinux reports: For some context here: originally, Proton had an option to enable / disable it globally. That was removed with the Game Recording update last year. That made sense, because people kept somehow turning it entirely off and now it's required by Steam. Currently, there's still an option in the stable Steam Client that you need to manually check to enable Steam Play (Proton) for "all other titles". This is something of a leftover from when Proton was initially revealed, and only worked for a specific set of games on Valve's whitelist. It now covers what Valve set by default for Steam Deck and SteamOS verification. What's changed is that at some point in the recent Steam Beta releases, is that "for all other titles" option is gone. I've scrolled back through changelogs and not seen it mentioned. So now, Proton is just enabled properly in full by default in the Steam Beta like shown in the [image here]. This is a good (and needed) change that I'm happy to see. There's often confusion when people try to run Windows games on Linux and end up with no install button because Proton isn't turned on for all titles. [This] will soon be a thing of the past. To be clear, this is not setting Proton on every game by default, it does not override Native Linux games. It's just making Proton available by default.

Read more of this story at Slashdot.

Probably the easiest way to a Google-free smartphone or tablet

Murena's /e/ OS is arguably the most mainstream de-Googled Android OS, and the new version catches up with important features.…

Startup has already shown how to make drugs in space

America's aviation regulator has issued its first license for unmanned spacecraft to reenter the Earth's atmosphere to Varda, a startup that's trying to build a space-based manufacturing business.…

Illegal products abound on Chinese site and its processes to stop ‘em are hopeless

The European Commission has found Chinese e-tail giant AliExpress in breach of its obligation to assess and mitigate risks related to the dissemination of illegal products as required under the Digital Services Act (DSA).…

The U.S. Army Reserve has directly commissioned four top Silicon Valley executives as lieutenant colonels under a new initiative, Detachment 201, aimed at accelerating tech integration into military operations. While these part-time roles are intended to bring private-sector innovation to defense modernization, the move is pretty unusual. Gizmodo reports: The Army said in a press release that the four executives are Shyam Sankar, CTO at Palantir; Andrew Bosworth, CTO at Meta; Kevin Weil, Chief Product Officer of OpenAI; and Bob McGrew, an advisor at Thinking Machines Lab and former Chief Research Officer for OpenAI. The four men are being commissioned at the high rank of lieutenant colonel as part of a program called Detachment 201: The Army's Executive Innovation Corps. As Task & Purpose notes, the men will get to skip the usual process of taking a Direct Commissioning Course at Fort Benning, Georgia, and they won't need to complete the Army Fitness Test. The Army didn't respond to questions emailed Tuesday but said in a statement published on its website that, "Their swearing-in is just the start of a bigger mission to inspire more tech pros to serve without leaving their careers, showing the next generation how to make a difference in uniform." Their role in the Army Reserve is to "work on targeted projects to help guide rapid and scalable tech solutions to complex problems," as the Army puts it. The new reservists will serve for about 120 hours a year, according to the Wall Street Journal, and will have a lot of flexibility to work remotely. They'll work on helping the Army acquire more commercial tech, though it's not clear how conflict-of-interest issues will be enforced, given the fact that the people all work for companies that would conceivably be selling their wares to the military. In theory, they won't be sharing information with their companies or "participating in projects that could provide them or their companies with financial gain," according to the Journal. Silicon Valley has always benefited greatly from ties to the U.S. military. Silicon Valley companies were bringing in $5 billion annually from defense contracts during the Reagan administration, something that the average person may not remember about the 1980s. But it's always been an uneasy alliance for consumer-facing tech companies, especially over recent decades. That's all changing, according to many folks who align more with President Donald Trump, who was once considered a shameful person to represent in polite company. As Andrew Bosworth, the CTO at Meta, who is joining the Army Reserves, told the Wall Street Journal, "There's a lot of patriotism that has been under the covers that I think is coming to light in the Valley."

Read more of this story at Slashdot.

Phishing, Python and RATs, oh my

A sneaky malware campaign slithers through Cloudflare tunnel subdomains to execute in-memory malicious code and give unknown attackers long-term access to pwned machines.…