AI Reading for Thursday, November 14

OpenAI told the all-hands that it plans to launch an ‘Operator’ AI autonomous agent feature in January as a research preview / beta test - The Verge

---

Another OpenAI safety leader quit - X (formerly Twitter)

---

Elon Musk's massive AI data center gets approved for 150MW of power from TVA, up from 8MW, enabling all 100,000 GPUs to run concurrently without (presumably illegal) gas turbines - Tom's Hardware

Definitely not crony capitalism.

---

Competitors flew spy planes over Musk's data center to see if his claims were for real. - The Verge

---

Insane AI slop about Musk swamps Facebook and other social media - 404 Media

---

Inside Physical Intelligence, a robot company from an all-star team from Berkeley, wants to be the OpenAI for robots - WIRED

---

Robots in China displace the lowest-skilled workers and force them to upskill or retire. - FT

---

IMF: next recession will cause an abrupt step function in AI automation - Fortune

---

Human translators will go the way of the Pony Express? - CNBC

---

CEO of DeepL, which has stolen a march on Google Translate, talked about real-time translation using LLMs. - Bloomberg

---

Homeland Security Department to Release New A.I. Guidance. - NY Times

The Crowdstrike incident highlights risks; one of the big cloud providers or AI providers pushes a bad update, has a major outage, a lot of stuff will stop working.

---

Microsoft ships industry vertical AI models for farms and factories. - VentureBeat

---

If online dating wasn't annoying enough, you are now probably getting catfished by an AI bot, excuse me, 'dating concierge' - NY Times

---

Unlocking the mysteries of complex biological systems with agentic AI and autonomous labs. If you want to breed plants with high sugar or THC, an automated lab with ability to cross-breed using reinforcement learning might be effective. - MIT Technology Review

---

A retrospective on Nvidia's rise and Jensen Huang's whiteboard - Barrons

---

More on Slack 'research' (cough - clickbait survey) that says management and underlings are misaligned on AI, workers are skeptical but may also use AI when it helps them keep above water, and hide it from managers - ZDNET

---

There's a boom in hiring experts to label data, eval responses, train AIs - Business Insider

---

At the end of this long conversation where someone was apparently working on homework, Gemini got fed up, and said, this is for you human, you are a burden on society, a stain on the universe, please die. AGI is officially here! - Gemini

Google Releases Standalone Gemini AI App for iPhone. You can now access misaligned easily jailbroken AI anywhere you go. - MacRumors

---

There is always a way to jailbreak, just like there is always a way to socially engineer someone. - IEEE Spectrum

---

Cursor bought Supermaven, which did some interesting stuff in code completion but didn't crack the top tier - TechCrunch

---

Using AI to help read all the scientific papers and conduct systematic research reviews. - Nature

---

Is AI behind a spike in insurance denials? - Newsweek

---

WSJ tests AI summaries - The Verge

---

The evolution of automation in the AI era: RIP RPA? - Andreessen Horowitz

You can get a lot of mileage from AI plus Zapier, Make, Automation Anywhere, or asking AI to write Python. I think future is no-code automation with tools like that. But huckster goofballs push expensive enterprise SaaS like UIpath and vertical solutions.

---

AI health care startup shutters after burning through $650m. In its most recent iteration it wanted to replace doctor visits wth automated CarePods. - Business Insider

---

AMD axes gaming, consumer jobs to invest in data center and AI - The Register

---

Perplexity launches ads - TechCrunch

---

Better living through AI (imaginary living, and more viral Instagram) - Tom's Guide

---

Follow the latest AI headlines via SkynetAndChill.com on Bluesky