Big News: OpenAI to Launch AI Model That Can Reason in 2 Weeks
A new rumor scoop by The Information, commented and analyzed by yours truly
OpenAI hasn't released anything worth of our time since GPT-4.
They've announced (and hinted to) many things in the year and a half since but we remain empty-handed. We glimpsed Sora. Nothing yet. Sam Altman and Jony Ive (designer of the iphone) discussed a new AI device. Haven't heard anything else. Altman promised Advanced Voice but most users are yet to get access.
OpenAI has given us signs at GPT-5, GPT Next (which is nothing but a placeholder), Orion… but nothing of substance. They've chosen to focus on releasing blog posts instead (an unusual strategy but let's not question Altman’s marketing genius).
Anyway, a new scoop by The Information suggests the slow, silent, under-the-radar phase for OpenAI is coming to an end (Jimmy Apples, OpenAI's number one leaker, is coming out of his patience cave, which can only mean something is coming).
The news is about Strawberry (formerly known as Q*), OpenAI's project to give AI the ability to reason. It's reportedly coming out as soon as in two weeks as a standalone offering in the ChatGPT website (possibly as a new model to choose from when chatting).
The Information shared four interesting data points that I analyze below. Remember, it's mostly unofficial comments by anonymous people in contact with (or inside of) OpenAI so there's no way to contrast the info until the official release.