AI Reading for Monday December 29

If your company is worth $4.5t, how much would you be worth if a paltry $125b didn't 'test investor faith' - the Guardian

Groq-blocking competition: Nvidia-Groq deal is structured to keep ‘fiction of competition alive,’ analyst says - CNBC

'nonexclusive' but I wonder what happens if e.g. AMD or Google wants to 'license' Groq's IP, presumably costs you $19b and you don't get the staff which might not not be a great deal. - TechSpot

The $20b loophole - DeepFabric

Nvidia said to plan its Feynman GPU architecture, successor to Rubin, to use TSMC A16 (1.6nm) hybrid-bonded 3D stacking of Groq-sourced LPUs and SRAM atop the GPU to accelerate deterministic LPU inference and reduce area and power. - techdeals.net

Big Tech gets everything it wants from the Trump administration, as it removed many A.I. chip export limits, fast-tracked U.S. data-center construction, backed crypto-related legislation, preempted state A.I. rules and allowed a stronger Nvidia chip sale to China. - The New York Times

Make AI great again?

Drain the digital swamp: DeSantis pauses Florida data-center projects, proposes deepfake and automation limits - Politico

Democrats (and Republicans) are split over AI as data-center protests spread, Gallup finds about 80% favor regulation and Pew reports 17% expect a positive long-term impact, while Anthropic CEO Dario Amodei warned AI could eliminate up to 50% of entry-level white-collar jobs. - Politico

AGI feels the Bern: Sen. Bernie Sanders called AI 'the most consequential technology,' urged study of mental-health effects and a moratorium on datacenters; Sen. Katie Britt introduced the Guard Act to ban AI companions for minors and impose criminal liability for harmful chatbot content. - the Guardian

OpenAI posted a job for a Head of Preparedness to track and prepare for frontier AI risks, build capability evaluations, threat models and mitigations, secure models before release of biological capabilities, and set guardrails for self-improving systems. - The Verge

my new title, director of doomsday prepping

In case of AGI, break glass: Do governments need a multilateral AI emergency response plan and a safety organization? - TIME

Governance by AI? AI tools are already embedded across government administrations, affecting about 30 million Americans via insurer and Medicare Advantage algorithmic care decisions, appearing in courts (Colombia 2023; U.S. judges 2025) and enabling AI-drafted laws (Brazil 2023). - Schneier on Security

Maybe an actual circuit could replace the 5th Circuit

All-Xi-ing Eye, made in USA: Chinese domestic-security agent used a U.S.-developed AI model to design a surveillance system targeting Uyghurs, labeled "Warning Model for High-Risk Uyghur Individuals," according to Jack Crovitz of Palantir Technologies and the Republic. - The Washington Post

XiGPT, AI with Chinese characteristics: China issued a draft regulating anthropomorphic AI, open for public comment through Jan 25, 2026, requiring clear AI identification, consent for training on people's data, bans on harmful content, two-hour usage reminders and human handover. - Gizmodo

Year of Qwen? - WIRED

an AI trained to pass 2000 ideological questions from Xi in your corporate Slack? Possibly trained with 'wake words' that trigger unexpected behaviors? Pass. You could use it for small classification tasks maybe.

A LocalLlama AMA with the Zhipu team - Reddit

Quick before Zee AI hallucinates and fires us

Modi's favorite billionaire launches AI center, urges India to build sovereign AI to protect jobs, data and national security - The Economic Times

digital pandit for making lit Bollywood deepfakes

A Russian embassy posted a cursed Christmas AI video - The Daily Beast

in Russia, Hallmark card say we will bury you

AI Chatbots Linked to Psychosis, Say Doctors - The Wall Street Journal

you think your computer is watching you and controlling your thoughts? crazy

FDA advisers call for evidence, transparency and postmarket monitoring for generative-AI mental-health chatbots - JD Supra

Tech companies have redefined 'friends' and 'connection' and are promoting AI 'companions' that are not genuine friends. - Financial Times

Mehdi Paryavi, CEO of the International Data Center Authority, warns excessive workplace AI use causes 'quiet cognitive erosion' and downskilling that undermines employees' confidence and critical thinking, per a Work AI Institute report with Notre Dame, Harvard and UCSB researchers. - Business Insider

my brain was quiet quitting before it was cool

Using ChatGPT to reframe nail-biting as anxiety- and ADHD-related, used AI check-ins and a fidget ring to reduce occurrences, and reported consumer LLMs could augment habit-change and mental-health apps without replacing therapy. - Tom's Guide

AI as the cause, and cure, for anxiety and all our problems.

Geoffrey Hinton warned that by 2026 AI, improving roughly twofold in task capacity every seven months, could replace many jobs, already linked to an estimated 30% drop in job openings since ChatGPT and greater wealth concentration. - Yahoo

Godfather of AI award moniker be a poisoned chalice

Commit message: deprecated the humans - AI researchers and economists say AI is increasingly automating code authoring—experts report AI contribution rising from about 10% to 75% for some engineers—and predict broad automation of many remote, computer-based tasks by around 2035. - Chris Barber

Sal Khan: Companies that deploy AI should dedicate 1% of profits to ease transition for those kicked to the curb by AI. - The New York Times

Speedrunning the white collar to blue-collar to prescription painkillers to early grave pipeline with the help of AI - The New York Times

The UK Association of Chartered Certified Accountants will scrap remote exams and revert to in-person testing after reporting AI-powered cheating outpacing safeguards, including one student allegedly photographing questions and feeding them to an AI. - Semafor

Michael S. Roth, president of Wesleyan University, says AI systems, Trump’s cabinet and higher education share a 'sycophancy epidemic' that erodes trust, threatens truth and democratic norms, and requires guardrails, meritocratic appointments and reformed AI governance. - The Washington Post

AI start-ups have amassed a record $150 billion funding cushion from mega rounds, creating "fortress balance sheets," while investors warn of bubble risks and advise top firms to brace for tougher markets and tighter follow-on financing. - Financial Times

Big Short 2: The Final Boss: Michael Burry Bets He Isn’t Too Early to Go Against the AI Juggernaut - The Wall Street Journal

Elon Musk has a long record of publicly setting aggressive deadlines his companies miss, and barring major breakthroughs, 2025 will likely add missed promises on robotics manufacturing, robotaxis and AI, eroding credibility. - The Information

At Least 15 People Have Died in Tesla Crashes Due to Faulty Electric Doors - Autoblog

China Is Banning Tesla-Style Retractable Door Handles Over Safety Concerns - Autoblog

When your door has one job and can’t do it because the world’s smartest man made it

Fox on steroids, with extra ‘roid rage: Viral racist AI-generated videos reshape public views on government programs and race - Axios

Kapwing found 104 of the first 500 YouTube Shorts (21%) in a fresh account were AI-generated, and ranked South Korea, Pakistan and the U.S. as the top three consumers, with the U.S. third by cumulative views (~3.39 billion). - PCMAG

Achy breaky AI heart: AI-generated country song 'Walk My Walk' by Breaking Rust topped a Billboard chart, prompting Nashville artists, songwriters and radio programmers to raise concerns about authenticity, chart integrity, royalties and calls for clearer labeling and industry rules for AI-produced music. - The Washington Post

Did Tim Cook post "AI slop" and lie about it? Or did an artist intentionally make something that looked like AI? And why is it a matter worth debating? - Slashdot Linux

Kling AI's motion-control tool automates complex motion tracking, generates or transforms footage via prompts, speeds VFX workflows and lowers filmmaking barriers while raising authenticity, provenance and reskilling concerns. - Reddit

Construction workers used AI-generated images to fake completed work, raising legal and reputational risks and likely increasing demand for verification and tighter insurer/platform vetting. - Reddit

too stupid to be real

“I just watched a machine make my entire Sweetgreen order.” AI slop, it's what's for lunch - TwistedSifter

WiseCode, founded in 2024, is a free AI-driven app that scans over 700,000 foods by barcode or photo, evaluates 15 attributes to produce a 1–100 WiseScore and uses proprietary FoodTechAI and datasets for ingredient and nutrient transparency. - CNET

Cat Olympic diving - Reddit

What an $11b data center looks like - Reddit

Ray Dalio says he's been using AI for 30 years. - X (formerly Twitter)

Well, I took an AI course in 1986, had nothing to do with AI as we know it now, was Lisp and expert systems with pattern matching and rules. Symbolics Lisp machine and interpreter were super cool tho.

Follow the latest AI headlines via SkynetAndChill.com on Bluesky