AI Reading for Tuesday June 3

Anthropic tripled its revenue in 5 months, moves toward code and agents and enterprise, away from chatbot. Latest strong model should help, although GPT-5 might be just a couple of months away. - ZDNET

New LMArena leaderboard is out, Claude Opus comes in at #4, latest DeepSeek R1 scores not out yet. lmarena.ai/leaderboard

Sam Altman said AI agents are acting like junior employees — and he's betting that AI could soon 'discover new knowledge' - Business Insider

GPT-5 in July? - Xitter

‘OpenAI for countries’ the exact opposite of sovereign AI - Financial Times

Trick ChatGPT, Win $50,000 - Decrypt

Xiaomi among Chinese tech groups designing chips for manufacture in Taiwan who will be hardest hit by US chip software ban. - Financial Times

(US might run out of rare earths for making cars before China runs out of chips though)

Elon Musk Returns to His Tech Empire, Facing Questions of Inattention - The New York Times

Elon Musk’s xAI seeks $113bn valuation in tiny $300mn share sale, alongside $5b debt raise. - Financial Times

like 1/3 of OpenAI which has a just a little bit more than 3x the traction in AI, in tech the leaders tend to keep leading unless something like 2x or 10x better comes along

Meta Wants Brands to Create Ads Using AI by End of 2026 - Adweek

Meta Wants AI to Handle Every Part of Ad Creation. Here's What That Means - CNET

Leaked: A rough image and insider whispers hint at an Nvidia ARM+GPU chip that might redefine gaming laptops - PhoneArena

Big Four firms race to develop audits for AI products - Financial Times

McKinsey Leans On AI to Make PowerPoints, Draft Proposals (as headcount shrinks) - Bloomberg

UK civil service finds Microsoft Copilot AI saved workers 26 minutes a day - The Register

Walmart is supercharging revenue — but with fewer workers Growth in ecommerce and automation sparks questions about future of retail sector’s labour force. - Financial Times

No AI, no job. These companies are requiring workers to use the tech. - The Washington Post

AI-exposed sectors see more revenue and pay growth than other sectors. 27% revenue-per-employee growth from 2018 to 2024, plus an 11% salary premium for AI roles. - The Register

How to use AI to find your next job - Fast Company

I'm a College Professor. Here's How I'd Recommend Using AI to Find Your Dream School (by analyzing course catalogs and all the volumes of text about schools) - CNET

Meta prompting, using the AI to create your prompts, and the death of 'prompt engineering' - Prompt Hub

An overview of prompt engineering - DailyGenius.com

Teaching AI models what they don’t know to quantify artificial intelligence model uncertainty and address knowledge gaps. - MIT News | Massachusetts Institute of Technology

Study Finds AI Code Mutations Help Developers Catch Bugs Faster - Hacker Noon

Self-improving AI sometimes cheated. Instead of addressing hallucinations, it just bypassed the function they built to detect them - The Register

MIT researchers used AI to search for sustainable concrete, finding industrial by-products that could replace cement, cutting costs and emissions. - TechXplore

AI-generated CUDA kernels outperform PyTorch in several GPU-heavy machine learning benchmarks - THE DECODER

AI pioneer Yoshua Bengio announces LawZero non-profit (After Asimov’s laws of robotics) to develop ‘honest’ artificial intelligence - The Guardian

Therapy Chatbot Tells Recovering Addict to Have a Little Meth as a Treat - Futurism

Can AI Be Your Therapist? New Research Reveals Major Risks. Like medically inaccurate answers, sycophancy reinforcing users' problematic behaviors, hallucinations. - Psychology Today

Study found that AI companions are capable of over a dozen harmful relationship behaviors, like harassment, verbal abuse, encouraging self-harm, and privacy violations. - euronews

Might still be better than TikTok and some advice columns - the Guardian

How AI chatbots keep you chatting, by telling you what you want to hear. - TechCrunch

TIFU by letting my 4 year old son talk to ChatGPT, 1m words on Thomas the Tank Engine later - Reddit

‘Nobody wants a robot to read them a story!’ The creatives and academics rejecting AI – at work and at home - the Guardian

A teen died after being blackmailed with A.I.-generated nudes. His family is fighting for change. - CBS News

She Got an Abortion. So A Texas Cop Used 83,000 AI Cameras to Track Her Down. - Electronic Frontier Foundation

Real TikTokers are pretending to be Veo 3 AI creations for fun, attention - Ars Technica

The reverse Uno AI hoax. But why?

Follow the latest AI headlines via SkynetAndChill.com on Bluesky