Is the AI That Got Sam Altman Fired Out? GPT-5?

There’s a new model on Lymsys.org called “anonymous chatbot,” apparently from OpenAI. I tried it, and it outperformed Gemini 1.5 Pro, the current leaderboard champion, on a simple prompt of mine. You can try it there too if you get it assigned randomly.

If the past is any guide, this testing is ahead of an important launch. Sam Altman posted this odd strawberry post. Strawberry is apparently the code name for Q*, which is rumored to be the AI Ilya saw before he fired Sam.

A lot of the subsequent OpenAI departures were due to concerns that OpenAI was moving too quickly without enough of a safety focus. Recent reports suggest that Strawberry is an advanced AI project focusing on deep research and enhanced reasoning capabilities. This project aims to enable AI models to autonomously navigate the internet, plan ahead, and solve complex problems, potentially bringing AI closer to human-like intelligence. These developments have sparked internal debates about balancing innovation with safety.

Combine that with the advanced voice and vision capabilities OpenAI is rolling out, and with Sora for image and video generation, it feels like this could be called ChatGPT 5 if OpenAI wanted to do so.

BTW — why does no one ask about safety at META, which has 10X the resources?