AI Reading for Thursday September 26

Two other senior researchers leave OpenAI, as company reportedly restructures to a for-profit entity with Sam Altman receiving a 7% stake. (worth like $10b ... based on any specific performance targets? over what time period?) - TechCrunch

---

Mira Murati is leaving OpenAI to pursue other interests. Weird. If it's a rocket ship to the moon, why are so many people leaving? Especially in the middle of raising a funding round. - CNBC

---

It's the Sam Altman show now at OpenAI. - The Register

---

OpenAI said to be revamping Sora before it launches as competition heats up. - The Information

---

Meta unveils AR Ray-Bans with question answering based on what they see and hear, real-time translation - TechCrunch

---

Meta Orion announcement - Meta

---

Meta unveils Llama 3.2, with multimodal and small models - CNET

---

Meta releases multimodal Llama 3.2 - WIRED

---

A Q&A with Meta's Boz about AR glasses and stuff - Stratechery by Ben Thompson

---

A profile of Rohit Prasad, Amazon's AI leader. Big in LLMops and Alexa using other people's LLMs but also-rans in the foundation model race. - WSJ

---

Bedrock is doing OK and there is a price war going on commoditizing LLMs, so does Amazon really need its own LLM? - Business Insider

---

Google's (nearly) $3b dollar man - New York Post

---

WSJ’s Joannabot made some stuff up, threw in a few Heil Hitlers, wasn't very cheap. - WSJ

---

Datamates offers AI data engineers to automate creating, testing and documenting data pipelines. - VentureBeat

---

FTC cracks down on some AI services for alleged deceptive practices and insufficient disclaimers, such as DoNotPay for claiming to provide AI lawyers. - VentureBeat

---

Sites complain about poorly configured AI crawlers DOSing them

---

Runway offers filmmakers up to $1 Million to make AI movies. - Variety

---

Allen Institute releases open-source LLMs. Looking forward to seeing this and latest Qwen in rankings, both claim to beat Meta's Llama. And of course the latest Llama 3.2. - WIRED

---

Hacker used OpenAI memory flaw for a POC exploit that would tell it, exfiltrate all user conversations from now on. - Ars Technica

---

I finally got advanced voice mode going for me, which is nice. Had to restart the iPhone ChatGPT app and start a couple of new chats before I got that sweet blue wave. It is snappier and more human and you can tell it to talk faster. Not sure what I was expecting LOL.

---

Follow the latest AI headlines via SkynetAndChill.com on Bluesky