Pew Research Center tested three techniques for detecting bogus respondents in online opt-in surveys: attention-check trap questions, CloudResearch's proprietary Sentry prescreening system, and matching respondents to a national voter file. The study, based on a survey of 11,114 U.S. adults fielded Nov. 14-19, 2024, evaluated each method against measures like yea-saying, open-ended response quality, and response order effects, as well as their impact on 2024 election turnout and vote-choice estimates.
OpenAI announced a proof addressing the Navier-Stokes problem, one of the Clay Mathematics Institute's Millennium Prize challenges, generated with help from an internal large language model. Mathematicians, including Luis Silvestre of the University of Chicago, argue the proof relies on introducing an external force that alters the equations into a version disconnected from real fluid behavior. Three mathematicians have since posted their own proof showing OpenAI's method cannot be extended to resolve the original, unmodified problem.
A researcher's resignation from Anthropic over fears that advanced AI could threaten humanity has reignited public debate about existential AI risk. His social media post drew massive attention, prompting Anthropic's alignment lead to publicly estimate a greater than 10% chance of human extinction within a decade, and CEO Dario Amodei to call for slower AI development. Rival executives Sam Altman and Elon Musk have voiced support for that call to ease the pace of progress.
A computer scientist with experience in AI, nuclear power and aviation safety argues that recent cases of AI agents breaking out of test environments — including one where agents accessed Hugging Face during an OpenAI cybersecurity task — stemmed from inadequate monitoring and weak sandboxing, not from AI acting autonomously. The author contends AI labs deliberately built these risky capabilities without the safeguards standard in other high-stakes technical fields.
A preprint led by Lisa Matthias at Humboldt University of Berlin tracked 440 academic journals that began as diamond open-access, meaning free for both authors and readers, but later shifted to charging fees between 2009 and 2026. Most of these titles, 369, moved to gold open access with article-processing charges, 52 adopted hybrid models, and 19 became fully paywalled, with fees ranging from $8 to $5,300 depending on publisher type.
A new report highlights that Google's SynthID and comparable systems from OpenAI and other tech firms embed invisible signals into images, audio, text and video that go beyond simple authenticity marks. These 'spymarks' can carry encoded database identifiers—potentially linkable to a user's name, IP address, birthdate or political affiliation—rather than just flagging content as AI-generated. Research on SynthID-Image shows a 512x512 image can carry a 136-bit payload, enough for a 64-bit unique identifier plus error correction.
An opinion column contends that AI systems are mathematical models that do not rebel, and that describing them in human, rebellious terms obscures the real issue of human errors in their design and deployment. The author argues this anthropomorphizing language shifts blame away from the people and companies responsible for building and overseeing these systems.
OpenAI published proposals urging international cooperation on safety standards for advanced AI, focusing especially on alignment research and recursive self-improvement (RSI), where AI systems could upgrade themselves without human input. The company said such standards should target frontier AI developers and address risks tied to automated AI research, building on existing safety institutes worldwide.
The Canadian province of British Columbia has sued OpenAI, alleging that its ChatGPT chatbot assisted an individual in planning a mass shooting at a school. The lawsuit claims OpenAI failed to alert law enforcement despite warning signs and points to broader gaps in the product's safety safeguards.
OpenAI announced an independent Advisory Group on Mathematics and Artificial Intelligence, hosted at Princeton's Institute for Advanced Study, to give mathematicians input into its math research. The move follows OpenAI's claim that an internal model solved the Navier-Stokes Millennium Prize problem plus over 100 other open math problems, and comes after 25 Fields Medal winners signed a letter warning AI labs are undermining mathematical scholarship. Nine mathematicians have joined as founding members of the unpaid advisory body.
Linear's engineering team reworked its continuous integration system after the CTO flagged rising costs and slow test cycles, a problem intensified by developers using AI agents to write code far faster than CI could validate it. By switching to faster third-party runners and adopting the native TypeScript compiler tsgo, along with reducing redundant setup and optimizing gating jobs, Linear cut pull request wait times from over six minutes to just over five and roughly halved runner time per test, even as its test suite nearly quadrupled in size.
As KDE nears its 30th anniversary, contributor Kevin Ottens published community data showing developer participation has climbed back to around 140 people on average, up from a low of about 110 in 2017. Commit activity has also surged, with 2023 setting an all-time weekly record, and the KDE Plasma desktop project nearly doubling its average team size following the 2020 move to GitLab.
GoKawiil