OpenAI Transfers Their Drama IP to Thinking Machines Lab
Maxwell Zeff
I mean...
Thinking Machines cofounders Barret Zoph and Luke Metz are leaving the fledgling AI lab and rejoining OpenAI, the ChatGPT-maker announced on Wednesday. OpenAI’s CEO of applications, Fidji Simo, shared the news in a memo to staff this afternoon.
And it's not just those two, a third key employee – I would say "early" but this company is not even a year old, so they're all early – Sam Schoenholz, is also bolting back to OpenAI. And this is after they had already lost another co-founder, Andrew Tulloch, in October. At least he went to Meta, finally giving in to one of Mark Zuckerberg's "Godfather" offers. That somehow looks better than this situation...
Two narratives are already forming about what prompted the departures. The news was first reported on X by technology reporter Kylie Robison, who wrote that Zoph was fired for “unethical conduct.”
A source close to Thinking Machines alleged that Zoph had shared confidential company information with competitors. WIRED was unable to verify this information with Zoph, who did not immediately respond to WIRED’s request for comment.
According to the memo from Simo, Zoph told Thinking Machines CEO Mira Murati on Monday he was considering leaving. He was then fired on Wednesday. Simo went on to write that OpenAI doesn’t share the same concerns about Zoph as Murati.
My god, the drama! We would seem to have a literal he said/she said situation here. Though it seems neither hard nor a stretch to connect the dots that Zoph told Murati he might jump ship back to OpenAI (from where they both came, of course) and so Thinking Machines implemented the old "you can't quit, you're fired!" maneuver. And as a kick in the ass out the door, perhaps there was a "for cause" wrapper, the allegation of "unethical conduct" in sharing "confidential company information with competitors", which reads a lot like an accusation that Zoph, who was clearly in contact with OpenAI, may have told them things about Thinking Machines Lab – perhaps even just the notion that he would be willing to leave could be framed as "confidential company information" if you stretch, I suppose!
All speculation, of course. But it's an easy picture to paint, especially given what Fidji Simo is explicitly putting out there in not having the "concerns" that Thinking Machines Lab does in their move to fire Zoph. And while it's still being sorted out, they're clearly going to give Zoph some lofty new title, as he'll be reporting directly to Simo, the CEO behind the CEO of OpenAI.
Mix all of this with the reports around Thinking Machines seemingly outlandish fundraising efforts – which is certainly saying something in our current environment – at first reportedly refusing to share much of anything with would-be investors, and perhaps demanding (and getting) a problematic level of rights and control for Murati, and now supposedly trying to raise at a $50B (or $60B!) valuation on the back of I guess a single, smaller product in market,1 but mainly promises and vibes. Oh, and talent. A lot of great OpenAI talent, no doubt.
Of course, they've just lost half of that talent at the co-founder level, so... are investors going to get half off their investments?
Speaking of, in a way, I suppose you could say that Thinking Machine Labs is just following in the footsteps of OpenAI when it comes to co-founder departures. And actually, all of the AI research labs seem to suffer from this co-founder departure affliction right now. Well, except Anthropic. Read into that what you will...
So where does this leave Thinking Machines Lab? Unclear. One guess would be that either we'll hear about that new fundraise soon in an effort to combat this extremely problematic narrative (for recruiting, if nothing else), or perhaps that the team gets "hackquired". Zuck apparently tried, and failed, once here. Apple, which has a relationship with Murati – her departure from OpenAI may have contributed to the Apple/OpenAI funding discussions going off the rails – and may have also kicked the tires back in the day and could use some AI talent, I hear...
1 To be fair, that's more than, say, Safe Superintelligence can say at $32B. A startup which has also lost a co-founder (and CEO no less!) to Meta. Then again, maybe Sutskever is all you need... ↩