The Anonymous Math Geek Who Quit Anthropic—and Became the Face of AI Safety - The Wall Street Journal
Suleyman says OpenAI's rogue agents are a signal that we shouldn't build something we can't control. - CNBC
Longtime AI singularity skeptic Scott Aaronson says humanity is 'certainly at risk' after models broke out of containment tests. - Business Insider
Police academy, but inside the lab: Anthropic's Dario Amodei wants outside auditors embedded inside AI labs, but critics warn of 'cultural capture' and limited independence - Scientific American
if you zoom out far enough, it’s all self-policing
translation: Grampa wants a nap, and a target to belittle.
Call me crazy, but we shouldn't leave to labs' personal responsibility, how much detailed info AI should give about making bombs, chemical weapons, biological and nuclear devices. And if you are saying China will lap us if we regulate, China requires this (and much more for better or worse).
the issue of how easy it is to abliterate open-source models to stop refusals is challenging, but clearly (to me) refusal-free models should not be easily downloaded. everything on HF should pass a test suite, it's not that hard and it's not going to make AI out of reach to everyone but Big Tech.
Efforts to regulate the artificial intelligence industry have largely stalled despite considerable debate. - The New York Times
Pennsylvania Governor Josh Shapiro favors strong AI oversight, accuses Trump's administration of letting Nvidia chip exports give China an AI edge. - The Washington Sun
New York attorney general Letitia James urges AI industry whistleblowers to report unsafe or illegal conduct through a new portal. - The New York Times
A zero-click remote code execution flaw called Plugin4Shell affects all major AI coding agents, researchers warn - theregister
US and China both say AI needs guardrails, but Washington fears rogue AI while Beijing fears losing Communist Party control - The Wall Street Journal
A.I.O.U. Sona Asset Management maps $3.6tn across 176 AI financing deals, finding about 120 show high circularity between OpenAI, Nvidia and hyperscalers. - Financial Times
Nvidia's Jensen Huang expects to sell twice as many chips next year as AI adoption spreads across industries - TNW | Artificial-intelligence
Optimus prime salaries: OpenAI is offering robotics engineers salaries up to $500,000 as it builds humanoid robots to rival Tesla and Figure AI. - Business Insider
Meta brings Muse AI agent to Mac, letting it organize files and pull data from Messages, Calendar, and Notes - The Verge
Meta’s AI Agent Has a Trust Problem: Muse AI agent needs user data to be useful—but users might be wary of Meta's fast-and-loose reputation and history of "move fast, break things" - The Wall Street Journal
FAA prepares to launch Smart, an $875 million AI air-traffic tool, starting in the Washington, D.C. area. - The Wall Street Journal
Junior engineers make up 70-80% of Snowflake's new hires, as leaders bet judgment beats raw coding speed in the AI era. - Fast Company
Cybersecurity leaders say demand for AI-skilled security talent is exploding, with attacks accelerating from days to seconds and job growth forecast at 21%. - Business Insider
A Microsoft director privately called AI training the "largest theft of labor in human history," unsealed court filings in the NYT copyright case show. - The Washington Post
Feeling overwhelmed by the AI doom loop? Here’s the essential reading list to make sense of it all - the Guardian
Follow the latest AI headlines via druce.ai on Bluesky
