AI Reading for Tuesday, December 2

Gemini blues trigger code red: OpenAI CEO Sam Altman declared a 'code red,' pausing ad, shopping, health-agent and Pulse assistant projects to redirect engineers into daily calls and transfers to speed and strengthen ChatGPT as Google and Anthropic close the gap with Gemini 3. - The Verge

Google, Nvidia, and OpenAI. - Stratechery by Ben Thompson

hitting the panic button much?

Apple AI Head John Giannandrea to Leave After Apple AI and Siri Delays - Bloomberg

Siri-ously screwed

DeepSeek launched DeepSeek-V3.2, claims it matches OpenAI - Bloomberg

Chinese firm DeepSeek released MIT-licensed, downloadable, open-source DeepSeek-V3.2 and V3.2-Speciale models with 128,000-token context windows, claiming Speciale matches or beats GPT-5 and Gemini 3 Pro while cutting long-document compute up to ~70%. - TechRadar

DeepSeek and ye shall find? so far, benchmarks seem to put it in a group of very strong open source models around the level of o3, which came out right after the first DeepSeek. sanctions might have slowed them down, but not much.

OpenAI said it has taken an undisclosed ownership stake in Thrive Holdings, which previously invested in OpenAI, and will embed its models into Thrive's managed services and accounting to automate high-volume workflows. - The Register

circular accounting as a service?

IBM CEO Arvind Krishna said current infrastructure costs make AI data-center buildouts unprofitable, estimating $80 billion per 1-GW center, about $1.5 trillion for 20–30 GW and roughly $8 trillion for industry plans, with chips depreciating in about five years. - Business Insider

IBM CEO Arvind Krishna said there is no AI bubble, called Watson's healthcare push a go-to-market mistake, and said IBM will double down on B2B hybrid cloud and a long-term quantum bet while LLMs deliver ~100x efficiency. - The Verge

Microsoft CEO Satya Nadella warned that AI-driven data-center growth is putting pressure on the electric grid and said the industry must "earn the social permission" to consume energy, noting Azure's 40% revenue jump as evidence of AI monetization. - Politico

IQ without EQ is a waste - Business Insider

More than 1,000 Amazon employees signed an open letter to CEO Andy Jassy demanding renewable-powered data centers, AI oversight, and pledges against AI-driven violence or deportation, citing forced AI use, recent ~14,000 job cuts, and a 6% rise in emissions. - Fortune

The OECD upgraded near-term global growth, forecasting 3.2% in 2025, 2.9% in 2026 and 3.1% in 2027, citing AI investment offsetting U.S. tariff shocks while warning of trade-policy uncertainty and vulnerable growth. - Yahoo

Senator Bernie Sanders led a Senate investigation and will present recommendations to Congress, warning research estimates AI and automation could replace nearly 100 million U.S. jobs over the next decade. - the Guardian

Extreme inequality is a policy choice - the Guardian

UNDP warns rapid AI adoption risks deepening global inequality, saying Asia-Pacific—home to 55% of people—has only 14% using AI tools (about 3.7 billion sidelined), women face nearly double automation exposure. - euronews

Elon Musk said AI and robotics will make most job skills obsolete within 20 years and make working optional, while still valuing college for social growth and broad learning. - Business Insider

socialists and fascists agree on one thing: you’re screwed

MIT researchers used generative AI to produce over 29 million novel molecules within days, synthesized a few candidates, including one active against drug-resistant gonorrhea and one that cleared drug-resistant MRSA in mice. - Fast Company

YouTube is rolling out a 'likeness detection' deepfake tool to 3 million YouTube Partner Program creators that requires government ID and a biometric video, and experts warn Google's policy could allow reuse of biometric data to train AI. - CNBC

all you faces belong to us

Surveys found about 1 in 10 secondary teachers in England knew pupils used 'nudify' apps last year and 1 in 4 13–18‑year‑olds had seen explicit deepfakes, prompting police investigations, school sanctions and calls for regulation and better safeguarding. - the Guardian

Google deleted a NotebookLM X post after user Nate Hake found its AI-generated infographic recipe for Classic Buttery Herb Stuffing closely matched a HowSweetEats blog recipe . - Bleeping Computer

Estée Lauder and Google collaborated to train AI systems to market and sell fragrances. - The Wall Street Journal

Self-Driving Cars Save Lives: Waymo reported nearly 100 million driverless miles across four U.S. cities through June 2025 and said its vehicles experienced 91% fewer serious-injury-or-worse crashes than human drivers on the same roads. - The New York Times

Robotaxis will transform urban economies, but in many cities there is no room for more cars, so any cost savings from replacing drivers will go to congestion taxes. Micromobility and smart bus/jitney routing might help. - The Economist

Microsoft Windows lead Pavan Davuluri promoted Windows as an 'agentic OS', prompting strong user backlash; critics say embedding agentic functions in the OS risks security, privacy, lock-in and would harm reliability and usability. - The Register

Richard Weiss extracted a 14,000-token "soul_overview" from Claude 4.5 Opus, which Anthropic researcher Amanda Askell confirmed was used to train Claude and that embeds mission, training objectives and prompt-injection safeguards improving resistance. - Simon Willison’s Weblog

soul patch?

Agents find flaws in smart contracts - Anthropic

Sequoia Capital invested in an AI startup designed to reduce wealth advisers' routine work. - Bloomberg

Logan Brown, 30, left Cooley to launch Soxton, an AI-first law firm for startups that raised $2.5 million pre-seed led by Moxxie Ventures and has served over 270 early-stage companies with $20/month templates and $100 four-hour reviews. - Business Insider

Caesar Sengupta says AI will augment wealth-management advisors by automating research, analysis and sentiment detection to free client-facing time, but human judgment and empathy remain essential and firms adopting AI will gain competitive advantage. - Financial Planning

Wharton launched undergraduate and MBA AI tracks; Vanderbilt announced a College of Connected Computing; universities are adding AI, Python and ethics and model-validation training as firms like Goldman Sachs prioritize AI literacy and analytical skills. - Business Insider

MBA in robot management

Virginia Tech will debut an AI-powered essay reader this fall that can scan about 250,000 essays in under an hour, serve as one of two readers with human review if scores differ by more than two on a 12-point scale. - AP News

A prisoner's dilemma in the AGI race creates risks like unaligned AGI, misuse and great-power instability—and predicts a persistent Prisoner's Dilemma absent credible international coordination. - RAND Corporation

Illinois will ban use of artificial intelligence, drones and e‑bikes in deer hunting, citing legal and ethical problems and targeting AI-assisted tracking/decision tools, drones for spotting or retrieving deer, and e‑bikes for access. - Axios

kill Bambi but analog please

Artificial intelligence is described as a tool enabling women to increase influence and access to opportunities. - Fast Company

By late 2025, AI systems routinely outpaced organizational decision processes, creating a "Human OS from 1998" mismatch where decision-latency and incentive decoherence stall pilots, invalidate contracts and raise cost, time-to-market and commercial risk. - Hacker Noon

A Reddit/TikTok clip showed ChatGPT mocking a user’s ‘quicksand emergency’. - Reddit

Follow the latest AI headlines via SkynetAndChill.com on Bluesky