Fired on Friday, former OpenAI CEO Sam Altman seemed for a moment over the weekend to be a hair’s breadth away from returning home like a victorious hero.
It would have been a new twist in an already dizzying saga. Mr. Altman held the cards: OpenAI employees rallied behind him upon his dismissal, and OpenAI shareholders urged the board to bring him back. Billions of dollars – and perhaps even the trajectory of the artificial intelligence industry – depended on the board, which seemed on the verge of bowing to pressure and backing down.
But no: the board, inflexible, rejected Mr. Altman’s return and affirmed Sunday evening in a memo to employees that his dismissal was “necessary” so that the board could “continue to fulfill its responsibilities and move this organization forward in its assignment “. Former Twitch boss Emmett Shear has been named interim CEO.
Shortly after, Microsoft CEO Satya Nadella announced the hiring of Altman and his right-hand man, Greg Brockman, to lead a new AI research division.
The OpenAI saga is far from over. The situation is fluid and there are still many unknowns, in particular the reason for the dismissal of Mr. Altman (the cause of which is not a specific event, but more generally a loss of confidence in him, indicated the board’s note on Sunday ).
But even without knowing the trigger, we can begin to assess the damage.
The big loser in all this is OpenAI.
Friday morning, the company was the absolute darling of techno, with its famous CEO, its flagship product ChatGPT, and its team at the cutting edge of AI envied by the giants of Silicon Valley. It was preparing financing that would have allowed employees to cash in their shares at a stunning price, and its language model, GPT-4, was at the cutting edge of AI.
OpenAI remains very dependent on Microsoft and its enormous computing power (to run its models) and which, since Monday, has within it a mini-OpenAI led by Altman and composed of defectors from OpenAI.
The OpenAI board may be happy with this outcome: they chose it – twice rather than once – despite the opportunity to reverse course. But he seems stupid for not explaining Altman’s firing. If he doesn’t provide more information, it’s hard to imagine employees accepting this decision.
No one had a more spectacular weekend than Mr. Nadella.
On Friday, when Mr. Altman was fired, Mr. Nadella seemed doomed to lose a crucial ally.
Mr. Nadella would clearly have preferred that OpenAI take over Mr. Altman. But the moment that idea was ruled out, he chose the best plan B: He offered Mr. Altman, Mr. Brockman and their loyalists a job on the spot.
From a tactical point of view, it is a masterstroke. In the short term, Microsoft may continue to use OpenAI models in its products. In the long term, Mr. Nadella gives Mr. Altman’s new team the money and computing power to create new models owned by Microsoft. He gets a group of exceptional AI researchers from OpenAI. As a bonus, Microsoft now owns 100% of a new AI lab that every Silicon Valley venture capitalist would have been willing to fund with their eyes closed.
For years, AI researchers and ordinary citizens have been warning that AI systems are becoming too powerful and that uncontrolled AI could pose an existential threat to humanity. These researchers are often affiliated with the “effective altruism” movement, which holds that reason and data can be used to determine how to act for the common good.
These people who fear the end of the world (the doomers) or demand a deceleration of AI research (the decels) were previously seen as outsiders. But their position is making headway in public opinion: numerous open letters are calling on governments to regulate AI. On Friday, they brought down the CEO of the world’s largest AI company.
Ilya Sutskever, OpenAI’s chief scientist and author of the hit against Altman, is not an effective altruist, but he seems to share their fears. Two of the board members who supported him, Tasha McCauley and Helen Toner, have ties to effective altruist groups.
If OpenAI suffers irreparable harm from the termination, the board will be blamed for destroying one of Silicon Valley’s most promising young companies and its billion-dollar value.
But the council, by virtue of its criteria, succeeded. He feared that Mr. Altman would rush headlong and build powerful and potentially dangerous AI systems, and he stopped him. It’s a victory for the cause, even if it comes at the expense of the company.
No one wanted Altman back more ardently than the investors and venture capitalists who stood to lose their money if he left.
These investors are now shareholders of a company without clear direction whose CEO is interim and the staff is in revolt. Worse yet, their only way to invest in Mr. Altman is to buy Microsoft stock.
Will OpenAI’s rivals benefit from Altman’s ouster? Hard to say.
Google, Anthropic and Meta could benefit from weakening OpenAI if they can catch up on AI or poach key personnel (as of Friday, recruiters were making offers to hapless OpenAI employees).
On the other hand, the net result is a stronger Microsoft. Moreover, Mr. Altman will no longer be hampered in his new initiatives by OpenAI’s hybrid governance structure (OpenAI Inc is an NPO, its subsidiary OpenAI Global is an ordinary company).
Which means Mr. Altman could go even faster.
