OpenAI’s convoluted corporate structure was designed to protect its board of directors from normal commercial pressures and ensure that the advanced artificial intelligence the company was building would always serve humanity.
Instead, the unusual arrangement tipped off a crisis this weekend that threatened to tear apart the company at the forefront of the global race over generative AI.
The news on Friday that four members of the company’s board had sacked chief executive Sam Altman shocked the tech world. Just minutes later, with no clear information about what had happened, one of the company’s investors claimed that this was how OpenAI was meant to operate — even if the repercussions might be devastating to a company whose value has soared this year to more than $80bn.
But as outside investors who had committed billions of dollars started to dig more into what had happened, shock turned first to dismay and then quickly to anger. None were presented with any clear reason for the ejection of an executive who had become the public face of the tech world’s AI boom.
The surprising sacking brought an immediate and remarkable groundswell of support for the ousted chief executive from senior tech figures and investors. By the next day, the backlash had hardened into a serious attempt by OpenAI’s investors, led by Microsoft, to reinstate him.
It was exactly the kind of outside pressure that OpenAI’s complex governance structure was designed to prevent. And by the end of the weekend, despite an expectation among some investors that the board would reverse its original decision, it looked as though OpenAI’s directors had refused to buckle.
In a dramatic move late on Sunday evening, the company picked Emmett Shear, co-founder of gaming service Twitch, as interim chief, leaving Altman out in the cold. By the end of the weekend, it would emerge that Altman and others loyal to him would instead move to Microsoft to set up a new AI division.
“There was no one big problem,” said someone with direct knowledge of the OpenAI board’s thinking. “The board reached the point where they couldn’t believe what Sam told them.”
Move fast and break things
The schism that opened up at OpenAI late last week owed much to the very different personalities and ambitions of the core group behind the company.
Altman was a restless and irrepressibly ambitious tech executive who had run Silicon Valley’s top start-up incubator, giving him a taste of how Silicon Valley could turn promising new tech ventures into world-leading companies almost overnight. He was closely allied with Greg Brockman, a former tech executive at fast-growing payments company Stripe, who became OpenAI’s chair.
Neither was steeped in deep learning, the technology behind the AI boom. That was the province of the third core member of OpenAI’s leadership group, Ilya Sutskever. A thoughtful Canadian computer scientist who was joint author of the formative paper that launched the deep learning era, Sutskever was more deliberative than the impetuous Altman.
He was also more prone in his public appearances to dwell on the potential dangers of the technology. Earlier this year he took joint leadership of a new group inside OpenAI trying to work out how to control future, superhuman AI systems. These, he wrote, could be “very dangerous, and could lead to the disempowerment of humanity or even human extinction”.
The stresses caused by OpenAI’s breakneck expansion this year added to those tensions, according to people familiar with the company, including over whether Altman was moving too quickly. Yet if this was one cause of the bust-up, investors and others say they have still not heard an explanation from the board of why Altman was fired so abruptly.
Altman’s ambitions had also spilled over beyond OpenAI, setting his interests in potential conflict with the company. One of his most audacious ideas, details of which emerged publicly for the first time this weekend, has been an attempt to raise tens of billions of dollars to create a new company to produce the chips needed to power OpenAI’s models.
Though it is still unclear what brought matters to a head, the crisis was triggered late on Thursday, when Sutskever told Altman the board wanted to talk to him at noon the following day. Only Mira Murati — the Albania-born chief technology officer and a close Altman lieutenant — was tipped off about the coming upheaval, since it would involve her stepping temporarily into the top job.
Joining Sutskever on the Friday video call, carried out over Google Meet, were the board’s three non-executives. One, Adam D’Angelo, co-founder of Quora and a former Facebook executive, was minor royalty in the Valley.
The others — safety researcher Helen Toner and tech entrepreneur Tasha McCauley — were less well known. Both were adherents of effective altruism, an intellectual movement that has gained backing in tech circles and holds, among other things, that AI could pose one of the greatest threats to the long-term flourishing of humanity.
Altman was sacked only minutes before the news was released. Brockman was stripped of the chairmanship and resigned within hours. It quickly became clear that his departure would rock the tech world, as well-known tech executives such as former Google chief Eric Schmidt and Airbnb co-founder Brian Chesky took to X to express support.
Yet by early the next day, there were signs that the abrupt move was being rethought, even inside OpenAI.
The terse announcement about the sacking said Altman had not been entirely open with the board. But in an email to staff, chief operating officer Brad Lightcap appeared to put an equal part of the blame on the company’s remaining four board directors, declaring: “This was a breakdown in communication between Sam and the board.” He also stressed that there had been no “malfeasance” or other serious business impropriety, giving Altman’s supporters greater confidence.
At the same time, investors had begun to put pressure on the company to reverse its decision and bring Altman back, with Microsoft and Thrive, the two biggest, in the lead.
Inside OpenAI, passions were rising. Staff members asked pointed questions about the departure at a hastily called internal meeting, according to reports of the meeting. By the end of Friday, three top engineers had resigned, raising worries about a much broader exodus. Some of OpenAI’s most prominent venture capital investors, such as Vinod Khosla, went public to say they would back any new AI company Altman launched, adding to the threat of mass departures.
Altman astutely played on the emotions. In a post on X late on Saturday he declared: “i love the openai team so much” — a message that brought a wave of heart emojis from staff, including some top executives.
“A vigil-like atmosphere” descended on OpenAI’s office in San Francisco’s gritty Mission district on Saturday, according to one investor. “Everyone was in but no one was working.”
Board digs in
With the ground shifting, OpenAI’s board did nothing to explain its abrupt actions. Some people involved in the effort to reverse Altman’s sacking expressed surprise over the weekend that the company had made no effort to build more support for its actions, either internally or externally, adding to the growing wave of support for bringing Altman back.
The outside pressure appeared to be making quick headway. Some pressing for Altman’s return believed it would be agreed by late Saturday, less than 36 hours after he had been ejected. But the negotiations stretched into Sunday, as the directors who had ousted him deliberated on their next steps.
The choice before them, according to one investor: they could stand by their original decision and risk an implosion at the company — and with it, the end of their long-term goal of building a safe AI for the world.
Or they could find a compromise that would put Altman back in charge and keep the mission alive, even if that led to an overhaul of the current board to ensure greater loyalty to the chief executive.
Some investors grew impatient as the weekend’s negotiations stretched late into Sunday, treating a reinstatement of Altman as almost a foregone conclusion.
But after an informal late-afternoon deadline for a decision came and went, reality sunk in: OpenAI’s board was not about to back down. At the end of Sunday night, Sutskever informed staff of Shear’s appointment, appearing to end hopes of a quick return for Altman.
But the company’s investors will continue to wrestle over its future. Microsoft’s chief Satya Nadella unveiled another bombshell at the end of a wild weekend of corporate intrigue.
While Microsoft remained “committed to our partnership with OpenAI”, he said that “together with colleagues, Sam Altman and Greg Brockman, will be joining Microsoft to lead a new advanced AI research team”.
A partner at one venture firm backing OpenAI was stunned by the weekend’s events: “I’ve never seen anything like this,” they said.
Read the full article here