Everybody’s out
garymarcus.substack.com
[resending to some readers, sorry for glitch]
Posted this a few minutes ago, giving my theory of what transpired over the last few hours:
and then, perhaps in keeping with my surmise that Sam’s bluff was called, negotiations broke down altogether.
Sam’s out, presumably Greg is too. Mira’s no longer interim CEO. Probably a lot of employees will leave. The $86B secondary sale to Thrive is presumably unlikely to close.
Big scoop from The Information:
The drama continues.
Gary Marcus never expected to spend this weekend quite this way.
When you're doing 100mph, your chances of surviving an impact plummet. The one thing I have not seen too many reporters focus on, in this story, is the ludicrous speed with which this is developing. To Gary's implied point at the very end, it's the weekend before Thanksgiving. The fact that the Board could not wait to let Sam go, speaks volumes. We're still waiting to hear what those volumes are actually saying.
Also, the fact that they all (Sam, Greg, the Board, the investors, etc) spent this entire weekend trying to resolve the situation by Monday (with two 5pm deadlines no less), speaks even bigger volumes.
This isn't FTX, but it strikes me that the more sudden and dramatic an implosion of a high profile company, the deeper and darker the core of the story.
Thanks Gary for keeping us updated
Erik Larson (author of the book "The Myth of AI") invited me for a guest blog on his substack: https://erikjlarson.substack.com/p/gerben-wierda-on-chatgpt-altman-and which was posted yesterday.
My guess there "[...] there is a conflict between the lofty goals — create safe and beneficial human-level AI for the world — of the non-profit OpenAI Inc. and the goals of the commercial sub-venture OpenAI LLC — which is there to make money, and which allows for commercial investors such as Microsoft to participate. It looks like Sam is looking for ways to escape the remaining constraints of the lofty non-profit. A new restructuring that turns the tables — the commercial sub venture becoming the lead and the not-for-profit lofty one simply becoming subsidised by the commercial arm — seems a possible outcome. Frankly, everybody is guessing now and so am I."
This turning of the tables *inside* the OpenAI structure seems to be off the table now. So, a new venture will be looked at. Now is the time we're going to see if it is possible to fund a second GPT-sized model (there is some uncertainty if there is a efficiency innovation in the wings that makes things cheaper — like retentive network for using or something else for training, otherwise I wonder how you are going to create a business case for training a second 500 million to 1 billion 'GPT-like' model and running it)