
Sam Altman reportedly said there are three possible endgames for AI safety: regulation, benevolent or extractive monopolies. - The Decoder

Race to the bottom…line: Dario Amodei defends Anthropic's "race to the top," citing nine costly safety decisions while admitting the company mishandled a draft AI bill - X (formerly Twitter)
Anthropic’s AI bill is weaker than the FRONTIER Act and makes Anthropic’s safety-first pitch ring hollow - Transformer News

Sundar Pichai says Gemini 4 Argon hit nearly 25 trillion tokens in 72 hours as Google cloud forgoes revenue to fund frontier AI - Big Technology
Pytorch will run but it’s a bit like releasing a new version of DirectX, the drivers are better on Nvidia, everything just works.

Hyperscaler 2027 data center spending forecasts near $1.1 trillion, but Bain sees AI revenue under $2 trillion by 2031 versus $6 trillion needed to justify investment - Advisor Perspectives
BlackRock CEO Larry Fink warns slow data center buildout would raise compute costs and be more inflationary - Bloomberg

Trump urges data center companies to pay local communities / NIMBYs a 'dividend' amid backlash that delayed or canceled 75+ projects - Business Insider
Edgerton, Kansas sued its own residents to advance an AI data center by a Trump-allied developer; residents seek a ballot ban - The Wall Street Journal

good for them but it’s all from renting their GPUs to Anthropic and others. anecdotally you can’t even obtain GPUs, 5090s were going for $10K until they got snatched off the market.
SpaceX shares rise 4% as revenue doubles on AI segment growth; Musk says SpaceX and Tesla formed committees to study a merger - The Wall Street Journal
AWS takes early ownership of about $5 billion in Nvidia GB300 racks, some stored in warehouses for lack of power - Business Insider
ByteDance, Alibaba and Tencent reportedly order over 3 million China-certified Nvidia RTX Pro 5500 GPUs worth $35 billion, far exceeding expected supply - Tom's Hardware



Competing AI systems for hospital billing and insurer claims review have added nearly $1 billion in costs, a Covered California official warns - San Francisco Chronicle
Jonathan Gluck, a cancer patient for 22 years, urges health insurers to drop AI agents handling claims and prior authorizations - Fast Company
OpenEvidence founder says two-thirds of US physicians use its ad-supported medical chatbot, sparking debate over patient care - Financial Times

Ethan Mollick charts 30 months of AI progress: 180 models beat GPT-4's launch, and AI now completes tasks taking humans 17 hours - One Useful Thing
Demis Hassabis expects AGI within two to three years, down from his three-to-five-year forecast earlier in 2026 - The Wall Street Journal
AI is sufficiently smart for many tasks and smarter models don't result in higher performance - X (formerly Twitter)


TripleTen data shows finance trails only tech in AI adoption: 48% of finance workers are asked to use chatbots versus 18% in healthcare - MIT Technology Review

Terminator TOS: UN disarmament institute will launch the first multi-stakeholder forum on military AI October 23, warning AI is crowding out human judgment in warfare - The Washington Post
Ukraine's front-line kill zone has become a testing ground for AI-enabled drones that autonomously identify and mark targets - The Wall Street Journal
Ministry of ungentlemanly agent warfare: MEK and UAE operatives built AI agents to impersonate activists, infiltrate 500+ Telegram groups and run hundreds of fake social accounts - Responsible Statecraft

Restaurants gonna love that
Target launches a beta Ask Target chatbot for the holidays, chasing Walmart, whose Sparky agent users build 40% larger baskets - Business Insider
Follow the latest AI headlines via druce.ai on Bluesky


