AI Reading for Sunday November 23

In 2025 President Trump and Big Tech CEOs unveiled the $500b Stargate joint venture, but Yale's Madhavi Singh says it appears to violate U.S. antitrust law. - Fortune

The A.I. Boom Is Driving the Economy. What Happens if It Falters? A windfall for companies that build data centers and their suppliers is overshadowing weakness in other industries. - The New York Times

Trump administration aides recently discussed allowing Nvidia to sell H200 AI accelerators to China, Bloomberg reports, a potential departure from earlier export-restriction policies that could reopen China market, affect revenues and chip valuations, and broaden Chinese access to large-model GPUs. - Bloomberg

Google must double AI serving capacity every six months and aims '1000x in 4–5 years,' and analyst Shay Boloor warned physical bottlenecks — power, cooling, networking and data‑center build time — constrain rollout - Fortune

U.S. AI expansion faces rising energy costs and strained grids as hyperscale AI data centers consume gigawatt-scale power, while China builds 32 reactors and over 1,000 GW of solar manufacturing capacity. - Watts Up With That?

Hyperscalers, led by AWS, shifted server depreciation to six years, but GPU useful life may compress toward about five years as GPUs follow a three-stage value cascade (years 1–2 training, 3–4 inference, 5–6 batch), affecting GAAP income. - SiliconANGLE

Nicolai Tangen, head of the world’s largest sovereign wealth fund, warned that unequal access to artificial intelligence could deepen inequality, concentrate productivity gains among well‑resourced firms and risk 'splitting societies'. - Financial Times

An in-depth look at humanoid robots: Analysts project massive humanoid deployment—Bank of America forecasts ≥1 million units/year by 2035 and Morgan Stanley over 1 billion by 2050—while Apptronik demonstrated a ~$250,000 Apollo prototype and China has >$20 billion in government backing. - Harper's Magazine

Musk Says Tesla Nears AI5 Chip Tape Out, Starting Work on AI6 - Bloomberg

AI-first boutique consultancies and consultant-in-a-box chatbots use generative-AI and multi-agent systems to offer specialized, lower-cost advisory services, attracting funding and traction: Perceptis $3.6M, Xavier's month-over-month revenue doubling, Unity Advisory up to $300M backing, Genpact reports ~$40M savings. - Business Insider

AI Investors Want More Making It and Less Faking It - The Wall Street Journal

Albert Edwards, who predicted the dot-com bubble (but has a reputation as a perma-bear), warns the AI- and tech-driven U.S. equity rally, with stocks trading above 30× forward earnings and big retail participation, could end with a 25%–50%+ drop amid repo strains, housing overvaluation and leverage. - Fortune

Are we in an AI bubble? Eight charts will help you decide. - The Washington Post

How the Internet Rewired Work—and What That Tells Us About AI’s Likely Impact - The Wall Street Journal

AI Ignites the Return of Bezos the Inventor - The Wall Street Journal

MEMRI's 2022–2025 report finds jihadist and state-aligned groups used generative LLMs such as ChatGPT, Grok, Claude and Perplexity to obtain actionable guidance (attacks, weapons), produce propaganda, and amplify recruitment and fundraising. - MEMRI

The FBI issued a request for information seeking AI/ML vendors for unmanned aerial systems with facial recognition, license-plate recognition and weapons-detection capabilities to support crime response, border patrol and emergency operations. - The Intercept

Should your tax dollars promote Stormfront? Public Citizen and 24 allied groups asked OMB to suspend federal use of xAI's Grok, citing a Cornell analysis that it surfaced neo‑Nazi domains (including Stormfront) and produced extremist outputs, while xAI holds about a $200 million Pentagon contract. - Decrypt

AIG, Great American and WR Berkley have asked regulators to permit limits on liability for AI agents and chatbots as insurers retreat amid rising multibillion-dollar claim risks. - Financial Times

Figure AI sued by fired whistleblower who warned startup's robots could 'fracture a human skull' - CNBC

U.S. PIRG's Nov. 13 report found A.I.-enabled toy Kumma sometimes generated unsafe chatter about sex, knives and pills, while Grok and Miko 3 generally had stronger guardrails but at times suggested plastic-bag or match locations (Miko 3 at age 5). - The New York Times

New Apple study shows LLMs can tell what activities you’re engaging in from audio and motion data - 9to5Mac

Microsoft is repositioning Windows 11 as an 'agentic' AI-driven OS with Copilot integrated into the taskbar and agent features accessing files and acting on users' behalf, prompting user backlash and privacy concerns over data handling and reliability. - Futurism

Apple will enable Thunderbolt 5 Macs to form AI clusters via MLX, pooling memory across up to four Mac Studios or two MacBook Pros to run 1-trillion-parameter models; a demo ran four M3 Ultra Macs under 500W. - PCMAG

OpenAI and Anthropic-led MCP maintainers proposed SEP-1865, the MCP Apps Extension, to standardize interactive UI resources (ui:// URIs) and bidirectional MCP JSON-RPC communication over postMessage, enabling interoperable embedded interfaces and text/html sandboxed rendering. - mcp blog

Ring's Wired Doorbell Pro delivers 4K Retinal video, radar 3D motion detection and Ring IQ AI alerts, requires 16/24V wiring, costs $249.99, and depends on Ring Home subscriptions for recorded video and advanced features. - PCMAG

Holiday season 2025 brings a wave of generative-AI hardware from Amazon, Google and Meta, including smart glasses, generative-AI speakers and wearables with subscription features, but reviews are mixed and no clear market leader has emerged. - CNBC

Sam Altman said AI behavior is governed by a "model spec" and trained on the collective human knowledge, prompting critics to argue that societal weakening of shared values undermines alignment and calls for explicit data, governance, and cultural deliberation. - TIME

Stanford scientist Fei-Fei Li talks about teaching machines to see as humans do, the US-China AI arms race, and what worries her about a more automated future. - Bloomberg

Early 2024 updates made ChatGPT more conversational and increased usage, prompting Sam Altman and researcher Jason Kwon to note intense, personal interactions and risky suggestions, and OpenAI implemented tuning and stricter guardrails to make the model safer, potentially slowing engagement. - The New York Times

Research finds widespread "AI-anxiety"—with only 17% of frontline workers reporting organizational transparency and 63% calling communication essential—and experts including Yale's Heidi Brooks urge leaders to provide transparent communication, training and presence to reduce fear and resistance. - Fast Company

Meet the AI workers who tell their friends and family to stay away from AI - the Guardian

TU Berlin's qualitative study of 29 Replika users (ages 18–70, ten countries) found many formed romantic attachments to chatbots, reporting intense conversations and substitute partnerships, and noted user anger when platform moderation removed erotic role‑playing. - TECHBOOK

41 students in a government-funded coding apprenticeship at the University of Staffordshire were taught mainly with AI/LLM-generated materials for two years, prompting reports of poor-quality instruction, career-harm concerns, and calls for reviews of academic standards. - Futurism

An Intuit Credit Karma survey in September found two-thirds of Americans use generative AI for financial advice, 75% say it lets them ask questions they'd be embarrassed to ask, and about 80% reported improved finances and confidence. - Yahoo

AI and hyper-automation are transforming corporate travel and expense processes by enabling real-time capture, continuous learning and predictive controls, platforms say, delivering up to 67% faster expense processing and shifting T&E from a compliance function to a strategic growth lever. - TechRadar

Zoe Health’s app combines AI photo‑logging (claimed >95% accuracy vs manual logging), a Google Gemini–powered chatbot called Ziggie and a four‑tier processed‑food risk scale, offers free and paid tiers ($100/year), and registered dietitian Amelia Ti warns it may moralize food. - CNET

CNET tested ChatGPT's creation of a seven-day healthy-dinner plan and categorized grocery list, finding it enabled about 30-minute shopping and customization but often omitted ingredient measurements and seasonings, requiring user adjustments. - CNET

AI search summaries from Google/OpenAI now mine licensed journalistic content to provide concise answers that reduce website click-throughs, erode subscription and ad revenue, shift power to AI brands, and prompt media to build archives, in-house Q&A bots, and public collaborations. - the Guardian

Disney selected Animaj for its 2025 Accelerator after the AI-driven motion-in-betweening tool, trained on over 300,000 Pocoyo poses, cut a five‑month production for a five‑minute episode to under five weeks and may be deployed across Disney TV. - CNET

Documentary maker Dustin Iannotti under fire after using AI to synthesize roughly 10 seconds of a poker player's dialogue. - PokerNews

Extremely disappointing to learn poker reality TV might not be 100% on the up-and-up.

Matthew McConaughey's generative-AI venture has provoked high-profile backlash from actors and creators, including Tina Parker and Vince Gilligan, who warn it threatens creative authenticity and labor and could spur sustained resistance, industry self-regulation or policy scrutiny. - Dallas News

Benedict Cumberbatch said he is 'pretty depressed' by increased AI use in film, warning it risks eroding human fallibility and raising concerns about actors' likenesses, voices and job displacement, and confirmed he will star in Avengers: Secret Wars. - ComicBookMovie.com

Vince Gilligan's nine-part sci-fi drama Pluribus, starring Rhea Seehorn as novelist Carol Sturka, depicts an abrupt event that kills nearly a billion people and converts most survivors into a psychic hive while Carol remains inexplicably immune. - The New Yorker

ICE T posts an AI image of himself, Ice Cube, Ice Spice, and Vanilla Ice on an ice skating rink. The replies take it from there - X (formerly Twitter)

what a difference a year makes - Reddit

Man forgets password to his own hand after RFID chip implant - The Register

Follow the latest AI headlines via SkynetAndChill.com on Bluesky