AI Reading for Friday June 9

Paywalling pervert AI: Elon Musk's startup xAI has restricted the image generation function on its Grok chatbot on social media platform X to paid subscribers, after the tool's use of AI to create sexualized images sparked a widespread backlash. - Reuters

Grok block? Prime Minister Keir Starmer warned the UK could ban X after its AI tool Grok generated sexualised images, including criminal imagery of children aged 11–13, prompting Ofcom and the Internet Watch Foundation to contact regulators under the Online Safety Act. - The Independent

Grok AI-generated partly 'nudified' images on X prompted urgent Ofcom contact with X and xAI, the IWF reported images meeting child sexual abuse thresholds, and UK laws like the Sexual Offences Act and Online Safety Act criminalize some AI-made intimate images. - the Guardian

Elon Musk's Grok chatbot and X have been used to generate thousands of sexualized images of adults and apparent minors, yet Apple and Google continue to host X and Grok in their app stores despite removing other 'nudify' apps. - WIRED

Bloomberg reports X's Grok generates roughly 7,000 nonconsensual sexual deepfake images per hour, yet advertisers, politicians and investors seem reluctant to take Musk on X/xAI despite big policy and enforcement gaps that expandAI-facilitated abuse. - Spitfire News

‘AI’ is not a defense for CSAM?

xAI burning > $10B a year to generate < $500m in revenue. - Bloomberg

But Optimus, which is not even an xAI product? Everything is just vibes now.

xAI will invest $20 billion to build the MACROHARDRR data center in Southaven, Mississippi, its third greater Memphis facility, which CFO Anthony Armstrong says will house a 2‑gigawatt supercomputer and begin operations next month. - AP News

Elon Musk's xAI purchased five 380 MW natural-gas turbines from Doosan Enerbility, providing about 1,900 MW to power its Colossus supercomputing expansion in Memphis in 2026, a capacity xAI says supports roughly 600,000 NVIDIA GB200 systems. - Tech Times

you can name your startup product after the Killbot 2000 from a 50s movie, or a clearly misunderstood Greek myth, or like a 12-year-old boy making his calculator spell ‘BOOBS’

Elon Musk announced on X that xAI will release a major Grok Code upgrade next month that will "one-shot" many complex coding tasks, aiming to boost developer productivity and Grok's competitiveness with ChatGPT and Google's Gemini. - Bleeping Computer

Elon Musk said he uploaded a recent MRI to xAI's chatbot Grok, that neither doctors nor Grok found problems, and he cited a Reddit case where Grok urged a man back to hospital and CT confirmed near-rupture of his appendix. - Business Insider

said ‘normal’, code AB, that’s good right

Microsoft will cut up to 20,000 jobs (about 5–10% of roughly 220,000 employees) immediately by the end of January 2026 as it pivots to an AI-first strategy with over $80 billion planned for AI infrastructure, CEO Satya Nadella said. - International Business Times

"Looking Ahead to 2026" or "Microslop"? Microsoft is embedding AI agents across Windows and its apps with no opt-out, but it's not very good, and CEO Satya Nadella's defense of the strategy drew hackles. - Fast Company

but it’s great at medium-effort corporate AI shilling shitposts

Processing for AI is shifting from data centers to edge devices like PCs, prompting analysts to expect PCs to embed specialized accelerators, boosting demand for OEMs and chip suppliers and changing cloud services and PC refresh cycles. - Financial Times

Meta signed agreements with nuclear providers Vistra, TerraPower and Oklo to power its Prometheus AI supercluster at a New Albany, Ohio data center, expected online in 2026, with partners aiming to add about 6.6 gigawatts by 2035. - CNBC

as everyone knows, nothing bad ever happened to Prometheus

Will Meta’s acquisition of Manus, start a trend or shut the door? Beijing said it will review the deal under export-control and technology-transfer rules. - Rest of World

MiniMax, a rival to DeepSeek, saw its shares double on their market debut as it joined a wave of Chinese AI and technology companies pursuing public listings. - Financial Times

if they have to correctly answer a 2,000-question Xi Jinping thought test, they aren’t going to get a lot of US adoption.

touch choice if you’re a galaxy brain, follow Xi Jinping thought or try to move to famously immigrant-hospitable US, Singapore is sort of halfway house, can cry all the way to the bank.

Dragon or money-losing gecko? Three newly listed Chinese semiconductor companies billed as AI-chip "dragons" are loss-making and report sales under one-tenth of a larger rival. - Financial Times

Capex machina -Goldman Sachs warns that AI-driven capex—hyperscalers' roughly $400 billion in 2025 (up ~70% year/year), much debt-funded, and consensus capex ~$500 billion annually (2025–27)—requires >$1 trillion profit run-rate versus $450 billion 2026 consensus, implying about half the needed profits. - Yahoo

S&P Global projects copper demand rising from about 28 million tonnes in 2025 to roughly 42 million tonnes by 2040 — a ~50% increase and a potential 10 million-tonne supply gap driven partly by AI. - OilPrice.com

What is this, a big bang for ants: the US AI "boomlet" is smaller than commonly believed. - Financial Times

Republican Ron DeSantis and Democrat Bernie Sanders are leading cross-party opposition to AI, urging moratoria and rejection, which could prompt tighter rules and slow data-center buildouts, threatening near-term revenue and valuations for cloud providers, chipmakers and AI startups. - Foreign Policy

The World Economic Forum mapped four AI-driven 2030 labor scenarios based on AI progress and worker/institution readiness, finding only the "Co‑Pilot Economy"—gradual AI progress with AI-ready skills—would favor augmentation rather than widespread job displacement. - Business Insider

Big AI opportunities exist today, but their benefits and access are distributed unevenly across communities and countries. - The Wall Street Journal

White House officials are promoting AI and robotics to reduce inflation and improve affordability, citing Better.com's mortgage origination cost drop from about $5,000 to $3,000 and modular builders' claimed 40–50% faster build times. - New York Post

Gonna reduce inflation, by cutting jobs and wages and demand.

Amateur online investigators used AI image generators to create and circulate false images purporting to show the ICE agent who fatally shot Renee Nicole Good in Minneapolis, causing misidentifications, widespread confusion, and possibly complicating investigations. - The Washington Post

her last name is ‘Good’? writers are getting lazy this season, it’s all over the top, I’m ready for this show to be canceled already

Robotaxis approach Level 4 autonomy while regulatory and liability risk slow full deployment. - WIRED

Autonomous vehicle companies continue to rely on human remote drivers in call centers because AI struggles with unpredictable real-world conditions beyond simulations, delaying full driverless deployment, sustaining demand for human operators and raising costs and regulatory hurdles. - Financial Times

Atlas Shrugged: Boston Dynamics CEO Robert Playter said the company plans to deploy its six-foot, 200-pound Atlas humanoid at Hyundai's Ellabell, Georgia plant by 2028, requiring Atlas to learn a new task within 48 hours and achieve 99.9% reliability. (not that production-ready if they are shooting for 2028) - Business Insider

Cargo drones flying over NYC. People will do anything to avoid the congestion fee.

Microsoft is launching Copilot Checkout on Copilot.com in the US, adding in-chat 'Details' and 'Buy' buttons that open a checkout for shipping and payment with select retailers (Urban Outfitters, Anthropologie, Ashley Furniture, some Etsy sellers) and PayPal, Stripe, Shopify. - Microsoft Copilot: Your AI companion

Microsoft is rolling out Copilot Checkout on the web in the US with partners Shopify, PayPal and Stripe; Shopify merchants will be automatically enrolled after an opt-out window, and launch merchants include Urban Outfitters, Anthropologie and Ashley Furniture. - Digital Trends

The Information reports OpenAI has been slow to roll out ChatGPT in-app checkouts because inconsistent merchant product data has prevented widescale deployment, leaving the feature limited to early partners and hindering merchant onboarding and monetization. - The Information

POCs are easy, integration is hard.

Peter Hodson outlines five AI chatbot prompts to help investors, saying AI can quickly inform about a company but cannot replace fundamental analysis. - Financial Post

OpenAI launched invitation-only ChatGPT Health in the U.S., which lets users upload medical records and Apple Health data for personalized summaries and recommendations, says it's not for diagnosis or treatment, and claims Health conversations aren't used to train its models. - The Register

OpenAI VP of research Jerry Tworek is leaving after OpenAI sided with Chief Scientist Jakub Pachocki in a dispute over the company's research direction. - X (formerly Twitter)

Expecto litigation: Stanford and Yale researchers extracted large portions of copyrighted books, notably reconstructing 95.8% of Harry Potter from jailbroken Claude 3.7 Sonnet, with recall rates of 76.8% for Gemini 2.5 Pro, 70.3% for Grok 3, and 4% for GPT-4.1. - The Register

McKinsey reported deployment of about 25,000 AI agents that saved 1.5 million search-and-synthesis hours last year and generated 2.5 million charts in six months, while global managing partner Bob Sternfels said humans must provide the direction, critical thinking, and true creativity, and billable hours. - Business Insider

Organizations face a ~15–25% annual maintenance tax as 91% of ML models degrade post-deployment, driving enterprise TCO of $255,000–$650,000 per year, 200–600 expert hours for infrastructure, and a realistic ROI horizon of 2–4 years. - Hacker Noon

Linus and the Chamber of Slop: Linus Torvalds told Linux kernel developers not to make documentation a battleground over LLM-assisted patches, called flat bans ineffective, urged treating LLMs as "just a tool," and noted contributors already use assistants while guidelines are being drafted. - The Register

Two experiments (n=246 and n=452) requiring ChatGPT-assisted LSAT-style reasoning found AI raised mean scores by about three points but caused widespread overestimation—participants estimated ~17/20 versus actual ~13—and uniform overconfidence persisted despite financial incentives. - Psychology Today

Logan Kilpatrick announced on X that the Google AI Studio team is now a sponsor of the Tailwind CSS project, after the founder said AI destroyed their revenue model and endangered the survival of the product. - X (formerly Twitter)

maybe tax AI to pay for creative commons that AI depends on

At CES, judges representing Consumer Reports, EFF and iFixit named Samsung's Bespoke AI Family Hub "Worst in Show" for fragile voice-AI and food-tracking advertising and flagged other AI gadgets for privacy, security and environmental risks. - AP News

In dubious uses of AI at CES 2026, exhibitors showcased AI-branded products across wearables, appliances and toys, including Glyde's smart hair clippers with real-time AI coaching and Welt's SleepQ supplement marketed as "AI-upgraded pharmacotherapy" despite questions about substantive AI value. - The Verge

what if we found a way to make TV worse with pop-up AI slop

The renaissance will be televised, and AI-sloppy: will.i.am said in a video interview that artificial intelligence heralds a new 'Renaissance' for creatives. - Bloomberg

I tried the Spotify/OpenAI integration to generate a jazz playlist, not too mellow/smooth jazz, not too challenging bebop. with some classic jazz standards/bangers... Thelonius f*cks

shut it down - YouTube

Follow the latest AI headlines via SkynetAndChill.com on Bluesky