How Openai scientist was wounded by failed coup
Sam Altman will keep a nervous eye on a ‘guiding light of our field’, writes James Titcomb
‘This is very sad to me; Ilya is easily one of the greatest minds of our generation and a dear friend’
When military coups fail, their protagonists are rarely sidelined immediately. Wary that they will be made martyrs, the lead rebels are generally kept around, even promoted, before they are later discreetly put out of the picture.
Silicon Valley, it would seem, employs the same strategy. Last November, Sam Altman and Greg Brockman, the chief executive and president of the Microsoft-backed artificial intelligence giant Openai, were deposed in a shock power grab.
The brief revolt, orchestrated by the company’s chief scientist and cofounder Ilya Sutskever, failed. Altman and Brockman were restored within days after staff threatened to quit en masse. The two returning executives apparently made up, with Sutskever, a Russian-born computer genius whom Altman called “a gem of a human being”.
But on Tuesday – almost six months to the day since the failed coup – that detente shifted again. Sutskever, a 38-year-old scientist, announced he was leaving Openai, saying he wanted to focus on “a project that is very personally meaningful to me”.
His departure caps a clear-out of potential challengers by Altman, who emerged from last November’s aborted coup with more power than ever over Chatgpt-maker Openai. However, Sutskever’s exit also potentially threatens to re-open the wounds that led to Altman’s dramatic ouster.
In choreographed statements on Tuesday, Sutskever, Altman and Brockman sought to emphasise that the split was amicable. “It was an honour and a privilege to have worked together, and I will miss everyone dearly,” Sutskever wrote.
Altman said: “This is very sad to me; Ilya is easily one of the greatest minds of our generation, a guiding light of our field, and a dear friend.”
The cosy semblance soon appeared to crack, however. After the news broke, Jan Leike, an Openai executive who had worked with Sutskever on developing safe artificial intelligence, revealed he had quit the company too.
Dispensing with the niceties of Sutskever’s statement, he simply tweeted, “I resigned.” The departures mean the two most senior executives at
Openai working on preventing AI’S biggest risks are now gone. In reality, Sutskever has been invisible for months. In January, when pressed for news about his whereabouts, Altman himself said he had no idea.
“I don’t know what the exact status of that is,” Openai’s chief executive said after being asked if his chief scientist was still working at the company.
Sutskever was born in Nizhny Novgorod in Russia, and his family emigrated to Israel when he was five as a generation of Jewish citizens left the collapsing Soviet Union. As an adolescent Sutskever enrolled in classes at Israel’s Open University, studying computer science courses remotely despite still being in high school.
After his family moved to Canada, he knocked on the door of Geoff Hinton, the celebrated British researcher known as the “Godfather of AI”, and asked for a job. Sutskever later studied at the University of Toronto under Hinton, working on breakthroughs that helped lead to today’s AI boom.
Sutskever followed Hinton to Google but in 2015 he received an email from Altman, inviting him to dinner with Elon Musk. Musk had fallen out with Google’s chief executive Larry Page
and was providing seed funding for a rival non-profit AI lab, which became Openai. Musk has described Sutskever as “the linchpin for Openai being successful”.
Sutskever and Altman shared a goal: developing “AGI” or artificial general intelligence, a system powerful enough to outdo humans at most tasks. Advocates say AGI will be the last human invention, because we will subsequently leave the task to machines. On the way to reaching this goal, Sutskever was instrumental in developing systems such as CHATGPT, which took the world by storm 18 months ago. While CHATGPT and Openai have now become globally famous, that success was the product of years toiling in relative obscurity. This work helped to bond the company’s original staff tightly. Sutskever officiated Brockman’s civil marriage ceremony in 2019, which took place at the company’s offices, underlining the closeness between Openai’s top brass.
However, in the months after Chatgpt’s release, Openai’s leadership became divided on the direction of the technology they had created.
Sutskever increasingly focused on fears that AI could become superhuman faster than we are able to tame it. He and other members of Openai’s board grew concerned about Altman’s apparent commercial bent and questioned his commitment to the company’s goal of developing AI safely.
On a fateful video call last November, Sutskever and board allies told Altman he had been sacked as chief executive, prompting an outcry from Openai’s staff and Microsoft, its key investor.
Within days, Sutskever shifted sides, saying “I deeply regret my participation in the board’s actions” and promising to reunite the company. Altman was restored as chief executive, while Sutskever was demoted from a newlook and more commercially minded board. While he remained at the company in the short term, his fate was potentially sealed, especially after a review of the debacle largely cleared Altman and restored him to Openai’s board. Since the coup, Sutskever is believed to have been a rare sighting at Openai’s San Francisco headquarters.
Altman has overseen a string of significant departures at Openai in recent months, helping to consolidate his power. His control over the company at the centre of the AI boom is now largely unchallenged – but he is likely to keep a nervous eye on what the genius behind many of Openai’s breakthroughs does next.