AI Reading for Saturday August 2

OpenAI raised $8.3B led by Dragoneer at a $300B valuation, months ahead of schedule and part of a $40B goal for 2025; 5M business users pay for ChatGPT - The New York Times

20b- and 120b-parameter OpenAI models spotted briefly on Hugging Face - X

Inside OpenAI’s rocky path to GPT-5: GPT-5 improvements won't be comparable to the leaps in performance of earlier models, such as between GPT-3 in 2020 and GPT-4 in 2023 - The Information

Leah Belsky, OpenAI's VP of education, said that students should learn to use AI as a tool to enhance critical thinking and creativity, not as an "answer machine," comparing AI to a calculator in education. - Business Insider

Anthropic Revokes OpenAI's Access to Claude Claiming TOS Violations as OpenAI tests GPT-5 - WIRED

Anthropic Revenue Pace Nears $5 Billion in Run-Up to Mega Round - The Information

Persona vectors: Monitoring and controlling character traits in language models - Anthropic

Anthropic studied what gives an AI system its ‘personality’ — and what makes it ‘evil’. The company can also tune the personality by injecting 'persona vectors' and is also hiring for an ‘AI psychiatry’ team. - The Verge

Robot psychiatrist, the hot job of the 21st century. Like a human who psychoanalyzes robots, and not a robot who counsels humans.

Ask Claude if it's conscious, it says it's not sure, unlike other LLMs that give a categorical no. - Scientific American

Google releases Gemini 2.5 Deep Think for AI Ultra subscribers - Ars Technica

Google rolls out Gemini Deep Think AI, a reasoning model that tests multiple ideas in parallel - TechCrunch

Apple Inc. Chief Executive Officer Tim Cook, holding a rare all-hands meeting following earnings results, rallied employees around the company’s artificial intelligence prospects and an “amazing” pipeline. - Bloomberg

A jury found Tesla 60% liable for a 2019 Autopilot crash, awarding $243 million in damages. - The Washington Post

Bill Gates says AI is moving at a speed that ‘surprises’ even him—and he says the experts can’t tell if it’ll replace humans in one year or ten - Yahoo

Big Tech's data center spending spree shows no signs of slowing - Business Insider

$344b this year for the top 5. - Bloomberg

Cerebras launches Code Pro ($50/month) and Code Max ($200/month), with Qwen3-Coder, running at speeds of up to 2,000 tokens per second, with a 131k-token context window, in your IDEs of choice, and no weekly limits. - Cerebras

Anaconda Python distribution with Jupyter and other tools is now the 'Anaconda AI Platform' - SiliconANGLE

Ubuntu is now the Ubuntu AI platform.

AI Threats Raise Demand for Cybersecurity Products That Don’t Exist (Yet) - The Information

AI agents could handle 80% of customer service tasks by 2030, reducing costs by 30% with human oversight. - Forbes

Can AI help solve the shortage of air traffic controllers? - The Washington Post

AI CT scans for heart patients - Nature

Using AI to monitor your child - CBS News

I'm a software engineer, and I've lost my job 4 times in the last 18 years. I don't think AI is the problem. - Business Insider

Well, when you put it that way, I believe you. Also, was told to use AI for work, pushes back, gets laid off, says it wasn't because of AI

Berkeley economist Brad DeLong says economic uncertainty is to blame for tough college grade job market, not AI - Fortune

Sprinters, Marathoners & Skeptics on the Future of AI & Power. - warontherocks.com

Not really in the sprinter camp, feels like 1 or 2 breakthroughs away from true AGI. I can't rule out that there is some Guns, Germs and Steel breakthrough out there that the AI leader might get to first and allow them to keep ahead til they screw up.

But you know, US got nukes and didn't immediately nuke everyone else, and a couple years head start only went so far.

Anyway the US isn’t gonna stay ahead if we shut down science and immigration, AGI is moot when you are going back to the 19th century, giving away leadership and committing national suicide. Don’t release stats the Great Leader disapproves of. Straight to jail, right away!

Hinton: Robots may invent their own language that humans can't understand - Business Insider

You really can't trust Zuck with collecting all your data as you walk around with AR glasses, he showed it repeatedly but this was Exhibit 1. - X

Shipped a network driver/proxy server to surreptitiously spy on users and competition and concealed that fact from e.g. Apple Store.

I made a notebook to demo the Model Context Protocol - GitHub

write a tiny MCP server and step through how it gets called.

Follow the latest AI headlines via SkynetAndChill.com on Bluesky