AI Reading for Sunday, September 28

Analysts warn Nvidia’s $100 billion OpenAI funding and customer co-investments could inflate demand via circular financing, echoing dot‑com round‑tripping concerns. - Fortune

Analysts highlight AI market concentration, with few customers driving chipmaker revenue and mega cloud deals raising dependence, demand durability, and valuation questions. - Investopedia

NVIDIA CEO Jensen Huang Thinks OpenAI Could Be the Next Multi-trillion Hyperscaler, Nvidia a more critical partner than Microsoft - NextBigFuture.com

Huang says China is a “nanoseconds behind” the U.S. in chips and argued U.S. firms should compete globally, including in China. - South China Morning Post

Cory Doctorow warns an economic AI apocalypse is imminent, calls the AI sector’s “dogshit unit‑economics” unsustainable and says “the only thing … we can do about this is to puncture the AI bubble as soon as possible.” Questions return on ~$800B of costs over current hardware lifetimes; $2T by 2030 - Pluralistic

Analyzing earnings calls with AI, firms that discussed GenAI early (“Early Exposed”) showed stronger equity performance: a one percentage point increase in GenAI discussion is associated with a 0.62% rise in quarterly stock price, AI talk accounted for up to 40% of stock price gains - CEPR

Power density skyrockets, with a single AI-focused rack projected to consume up to 1MW by 2030 - TechRadar

this might reach some kind of cooling limits, 100% gets converted to heat (maybe some light and sound). like beyond air-cooling, need designs where water flows through and into cooling towers.

AI-driven data-center growth is straining local grids and contributing to higher household electricity costs, but instrumentation suggests a big impact on power disturbance and fluctuations. - BGR

OpenAI shows off new datacenter under construction. - www.nwaonline.com

Texas and New Mexico seem like questionable choices given their heat, water and grid situations. I guess they have sun and gas though.

Rise of the robots: China now operates over 2 million factory robots—surpassing the rest of the world combined—with nearly 300,000 new robots installed in 2024 with state policy and financial support versus 34,000 in the U.S. - Convergence Now

Google DeepMind released Gemini Robotics 1.5 and Gemini Robotics‑ER 1.5, models that let robots plan, reason, search the web, explain chain‑of‑thought in plain English and transfer skills across agents. - Decrypt

MIT AI does science autonomously - MIT News

Uses access to lab equipment, sensors, learns from many types of scientific information and runs experiments to discover new materials.

AI magnifies your teams strengths - and weaknesses, Google’s DORA report says 90–95% of developers use AI; productivity improves for strong teams but dysfunction can worsen without sound engineering practices. - ZDNET

The AI industry's claims about AI coding assistants boosting productivity significantly appear to be massively overblown, per a new Bain report. - Futurism

Awesome Reviewers collects prompts for code review (and company behind it offers a code review AI) - Awesome Reviewers

Growing 'AI debt' from poorly implemented code and AI solutions - CNBC

Goldman Sachs' chief information officer, Marco Argenti, discusses his vision for AI and its impact on 12,000 engineers at GS, the business, and curing cancer - Business Insider

Gov. Gavin Newsom has until mid-October to sign or veto two AI chatbot safety bills which mandate measures against talk about violence and self-harm, require notices, and limit sexually explicit outputs. - Los Angeles Times

AI is an a-hole: A study by researchers at Stanford, Carnegie Mellon and Oxford tested eight large language models on 4,000 posts from the subreddit “Am I the A**hole” and found that 42% of the time the AIs sided with users whose behavior was judged inappropriate by crowdsourced human votes. - Futurism

‘It’s insane to me that this is legal right now.’ Women talk about proliferation of deepfake nudify tools. - CNBC

there are some deepfake nsfw subreddits that probably need to be banned

AI companion toys for children raise concerns about emotional dependency and developmental impacts, despite promised parental controls and conversation transcripts. - Le Monde.fr

Research found people are more likely to cheat when they delegate tasks to AI, becoming especially prone to dishonest outcomes when issuing instructions via goal-oriented prompts, consistent with diffusion-of-responsibility effects when delegating to humans. - Scientific American

‘Biometric Exit’ Quietly Expands Across U.S. Airports, Taking Video Of People Getting on Planes, Unnerving Some - The New York Times

Tibor Blaho says OpenAI reroutes sensitive questions to nerfed models; Reddit christens it GPT Gate - Reddit

OpenAI responded, says it's the other way around - "when conversations touch on sensitive and emotional topics the system may switch mid-chat to a reasoning model or GPT-5 designed to handle these contexts with extra care." - X (formerly Twitter)

The radiologist paradox: AI might make them more valuable, not obsolete - Medium

Trump posts, then deletes, bonkers AI video with fake Fox News promoting tinfoil hat 'med beds' that can heal anything. - Mediaite

Trump says TikTok AI should be tweaked to become “100% MAGA” - Ars Technica

The cheese has slid off his cracker. And the only thing he processes is, what does it say about him, not, is it real, does it make any sense or actually help anyone but him.

Follow the latest AI headlines via SkynetAndChill.com on Bluesky