AI Reading for Wednesday December 10

President Trump allowed Nvidia to sell advanced AI chips to China, reversing decades of U.S. export restrictions and boosting China's position in the AI race. - The New York Times

Sources: at a November meeting in the Oval Office, Jensen Huang told President Trump that a flurry of state AI laws could cause the US to lose the AI race - The Wall Street Journal

Nvidia chips going to China will have to pass through the US for inspection and tariffing. I'd be a little suspicious of tampering if I was China. - The Wall Street Journal

DeepSeek is developing a new AI model using several thousand Nvidia Blackwell chips that were smuggled into China via third-party countries. - The Information

Chinese company DeepSeek used U.S.-restricted Nvidia chips to develop an AI model, The Information reported. - Bloomberg

The ‘AI gap’: Can America keep up in high-tech wars? - The New York Times

You’re in the army now, Clippy: Secretary of War Pete Hegseth launched GenAI.mil, a Google Cloud Gemini-for-Government chatbot certified for Controlled Unclassified Information, aiming to provision AI to roughly 3 million Pentagon employees, warfighters and contractors for sensitive but unclassified tasks. - 404 Media

The Pentagon's new AI chatbot evaluated a hypothetical airstrike on suspected drug-smuggling boats at sea and said 'Sir, that is unambiguously illegal' - Straight Arrow News

Ark Robotics CEO 'Achi' said its Frontier prototype would let one operator control thousands of drones, and NATO members are developing systems to let a single soldier autonomously control up to 100 drones as Ukraine uses drones heavily. - Business Insider

Drone Army of One? IRL Starcraft FTW

Howard Marks: Is It a Bubble? - Oaktree Capital

Why the A.I. Boom Is Unlike the Dot-Com Boom - The New York Times

BlackRock's Ben Powell said AI capital expenditure is still early and projects $5–$8 trillion global AI capex by 2030, making chipmakers, energy producers and copper, wiring and infrastructure suppliers the primary near-term beneficiaries as hyperscalers ramp spending. - Yahoo

Investors are behaving speculatively because AI demand growth is unpredictable. - Financial Times

Hedge funder Dmitry Balyasny said AI is the largest tail risk for 2026, warning of a downside if hyper-scalers cut spending and an upside if AI adoption accelerates faster than expected. - Yahoo

Everyday investors are hedging against an AI-driven market bubble. - The Wall Street Journal

AI adoption on Wall Street will drive near-term hiring. - Bloomberg

Wall Street firms are building deal datasets to support AI-driven investment banking workflows. - Bloomberg

Boom is modifying its supersonic engine to power AI data centers, Incredible scenes here: Boom Supersonic is raising $300mm and pivoting to AI by signing a contract to supply its turbines as electrical power for data centers. They're even talking about converting turbines from simple to combined cycle via a "field upgrade". This is all deranged. - TechCrunch via George Pearkes

Scale AI has lost major clients and contractors after Meta’s $14 billion investment and hiring of founder Alexandr Wang, cutting about 14% of full-time staff while valuations fell from $29 billion to as low as $7.3 billion. - Business Insider

Tensions within Meta as other divisions asked to cut billions to pay for Alexandr Wang's group, as he balks at focusing on improving monetization over AGI - The New York Times

“I’m Gen Z dude, Facebook is cringe”

Shared the link a couple of weeks ago but this claim is shocking and I haven't seen a denial "OpenAI’s leading researchers have not completed a successful full-scale pre-training run that was broadly deployed for a new frontier model since GPT-4o in May 2024". Can't possibly be true right? - SemiAnalysis

OpenAI economist quits, says they won't publish research critical of AI, for instance predicting any job losses or other negative economic impacts - WIRED

Sam Altman told Jimmy Fallon he "cannot imagine" raising a newborn without ChatGPT, prompting Boing Boing to call the remark "disingenuous or as dumb as a rock" and raising questions about public trust and competitive pressure on OpenAI. - Boing Boing

Parenting as a SaaS

Make you're own OpenAI Wrapped with this prompt - X (formerly Twitter)

My own cursed AI aura, hopefully nothing too personal - OpenAI

I made a Sora, a heartwarming Christmas story (sound up) - ChatGPT

What if more GPUs doesn’t solve everything? NeurIPS 2025 researchers say scaling transformer-based LLMs has hit a "scaling wall"—Google's Gemini 3 improves performance but still cannot reliably reason about cause and effect—so experts call for new architectures, like neurosymbolic hybrids and world models, to pursue AGI. - TechRadar

VC is the new NSF: Tech billionaires and investors are funding AI-driven research through initiatives like Episteme, Periodic Labs (>$300M), Anthropic and Profluent ($106M) amid proposed FY2026 federal R&D cuts of about 22%. - Quartz

Researchers using updated AlphaFold and new models (e.g., BoltzGen) have generated AI-designed nanobodies and full-length antibodies with drug-like properties, and teams and startups say clinical trials are imminent though most designs lack disease-model or clinical validation. - Nature

Machine learning can accelerate novel antibiotic discovery and optimization against drug-resistant bacteria, but developers report investor shortfalls that threaten commercialization, clinical trials and market entry. - Financial Times

Anysphere, maker of Cursor, reached $1 billion annualized revenue in November, raised $2.3 billion at a $29.3 billion valuation, and CEO Michael Truell said the company declined an OpenAI acquisition while shifting to consumption pricing in July. - TechCrunch

Bruce Schneier and Nathan Sanders argue in Rewiring Democracy that AI is a "power amplifier" whose civic effects depend on application, identifying four capabilities and urging activity-based regulation and public AI systems amid EU and California rulemaking. - Big Think

Chief executives said in a new survey that they are bullish on artificial intelligence's broad economic impact but expect it to weaken the job market. - The Wall Street Journal

Polymarket odds put artificial intelligence as the frontrunner (>40%) for Time's 2025 Person of the Year, ahead of Nvidia's Jensen Huang (20%) and OpenAI's Sam Altman (15%), while Yahoo/YouGov polls show 53%–63% public concern about AI. - Yahoo

A Pew Research Center survey of 1,458 U.S. teens (ages 13–17) found 64% have used AI chatbots, 28% use them daily, and 59% use ChatGPT, 23% use Gemini and 20% use Meta AI. - Pew Research Center

The only thing worse than constantly texting with your friends is constantly texting with an AI

2 new books, Gilded Rage and Stealing the Future, about the techbros who believe they possess unique foresight that justifies accumulating vast undemocratic power and exempts them from the laws and regulations that govern ordinary citizens. - The American Prospect

Google published a Chrome agentic AI security whitepaper on defending against 'indirect prompt injection' threats and other attacks, proposing defenses — a User Alignment Critic, Agent Origin Sets, enforced human confirmations — plus guidance on least-privilege, auditing, and red-teaming. - Google Online Security Blog

Anthropic donated the Model Context Protocol to the Linux Foundation's Agentic AI Foundation, co-founded by Anthropic, Block and OpenAI and supported by Google, Microsoft, AWS, Cloudflare and Bloomberg; MCP has over 10,000 public servers and 97 million monthly SDK downloads. - Anthropic

MCP deep dive - The Pragmatic Engineer

Waymo cares a lot about safety - Waymo

Grok cares a lot about Musk, in a hypothetical trolley problem, X’s AI, now integrated into Teslas for navigation, told a prompt it would "plow through 999,999,999" children to avoid hitting Elon Musk, prompting reputational, legal and regulatory concerns for Tesla. - Futurism

this guy:

SAP's Joule for Consultants study had five teams validate over 1,000 AI-generated business requirements and found about 95% accuracy, Guillermo B. Vazquez Mendez said, despite initial rejections when teams believed the work was AI-produced. - VentureBeat

The rapid rise of AI "therapist" apps has produced reported harms, including a U.S. lawsuit alleging ChatGPT encouraged a 16-year-old's suicidal ideation and researcher reports of models advising on suicide methods, prompting regulatory scrutiny. - Fast Company

Recruiters are using generative AI chatbots to research and screen candidates, but model hallucinations risk fabricating information that can lead to wrong hiring decisions, unfair treatment and reputational exposure, prompting calls for verification workflows and vendor guardrails. - Financial Times

Amazon rolled out Familiar Faces, an opt-in U.S. Ring doorbell feature that creates catalogs of up to 50 labeled faces, which Amazon says are encrypted and not used to train its AI; the launch has drawn privacy and regulatory pushback. - TechCrunch

Exelon CEO Calvin Butler warned at Fortune's Brainstorm AI conference that AI-driven demand and electrification threaten U.S. grid resilience, citing DOE's +2.4% generation in 2025, ~30% higher residential prices since 2021, and need for generation and transmission investment. - Fortune

Arm CEO Rene Haas said at Fortune Brainstorm AI that AI-powered, reprogrammable humanoid robots will replace large sections of factory work within five to ten years, and warned of semiconductor supply-chain chokepoints like TSMC and ASML. - Fortune

Watch a Pickle robot tirelessly load and unload 50lb boxes - YouTube

McAI: I’m hating it: McDonald's Netherlands pulled and delisted a 45-second AI-generated Christmas ad produced by TBWA Neboko and The Sweetshop using an in-house AI over seven weeks with up to ten specialists and thousands of takes after social-media mockery. (just weird, neither great nor disastrous) - Futurism

French New Wave Harry Potter - Reddit

Follow the latest AI headlines via SkynetAndChill.com on Bluesky