Enlarge / Ilya Sutskever, OpenAI chief scientist, speaks at Tel Aviv University on June 5, 2023.
On Friday, OpenAI unexpectedly fired CEO Sam Altman, leading to the resignation of President Greg Brockman and three senior scientists. The move also surprised major investor and minority owner Microsoft and reportedly infuriated CEO Satya Nadella. As Friday evening progressed, reports emerged that the fall was likely orchestrated by chief scientist Ilya Sutskever due to concerns about the security and speed of OpenAI’s technology deployment.
“In doing so, the board fulfilled its duty to the nonprofit’s mission, which is to ensure that OpenAI builds an AGI that benefits all humanity,” Sutskever told staff at an emergency all-hands meeting Friday afternoon, The Information reported .
Since its inception, OpenAI has pursued the development of artificial general intelligence (or AGI), a hypothetical technology that would be able to perform any intellectual task performed by a human, potentially replacing large numbers of humans in their workplace.
Internally at OpenAI, insiders say disagreements arose over the speed at which Altman pushed for commercialization and company growth, with Sutskever advocating for slowing things down. Sources told reporter Kara Swisher that OpenAI’s Nov. 6 Dev Day event, where Sam took center stage in a keynote touting consumer-like products, was a “tipping point for Altman pushing too far, too fast.” “.
In a joint statement released Friday evening, Altman and Brockman said they were “shocked and saddened” by the board’s actions. And they weren’t the only ones shocked by the news, as tech insiders took to social media on Friday to share their reactions. Angel investor Ron Conway wrote: “What happened today at OpenAI is a boardroom coup the likes of which we haven’t seen since 1985, when then-Apple CEO ousted Steve Jobs. It is shocking; it is irresponsible; and that is not the case.” directly from Sam & Greg or any developer in OpenAI.
OpenAI has an unusual structure in that its for-profit arm is owned and controlled by a 501(c)(3) nonprofit public charity. Before yesterday, this nonprofit was controlled by a board that included Altman, Brockman, Ilya Sutskever, and three others who were not OpenAI employees: Adam D’Angelo, the CEO of Quora; Tasha McCauley, associate senior management scholar at RAND Corporation; and Helen Toner, director of strategy and basic research grants at the Georgetown Center for Security and Emerging Technology. Now only Sutskever, D’Angelo, McCauley and Toner remain.
Enlarge / A block diagram of OpenAI’s unusual structure, provided by OpenAI.
Surprise movements and commotion
According to the joint statement from Brockman and Altman, Altman’s firing came as a complete surprise to the two and they laid out a rough timeline of what happened. On Thursday evening, Altman was asked to attend a remote board meeting at noon on Friday. The next day, Brockman, the CEO of OpenAI, was not invited to that board meeting, after which Altman was fired.
About 30 minutes later, Brockman was informed by Sutskever that he was being removed from his board position but could remain with the company, and that Altman had been fired (Brockman declined and resigned his position later on Friday). According to Brockman, the OpenAI management team was only made aware of these moves shortly after the incident, but former CTO (now interim CEO) Mira Murati was informed on Thursday evening.
Important questions remain about the allegations made against Altman in the OpenAI blog post announcing his resignation. In it, the board said Altman “was not consistently open in his communications with the board, which hindered its ability to carry out its responsibilities.” That has yet to be clarified by the company, but insiders say the move was largely a power play that resulted from a cultural divide between Altman and Sutskever over Altman’s management style and his drive for high-profile publicity. On September 29, Sutskever tweeted: “Ego is the enemy of growth.”
Enlarge / Sam Altman and Ilya Sutskever speak together at Tel Aviv University on June 5, 2023.
The schism causes further turmoil within. As noted, three AI researchers loyal to Altman also left the company on Friday and resigned in response to the news: Jakub Pachocki, GPT-4 lead and OpenAI research director; Aleksander Madry, leader of an AI risk assessment team, and Szymon Sidor, an open source baselines researcher.
Pushing back the “veil of ignorance.”
Rumors are already circulating about possible internal breakthroughs at OpenAI, which may have exacerbated the slow/fast divide within the company, due to Sutskever’s role as co-leader of a “superalignment” team tasked with figuring out how a hypothetical superintelligent AI can be controlled. At the APEC CEO Summit on Thursday, Altman said, “Four times in the history of OpenAI – the last time just in the last few weeks – I have been privileged to be there as we lift the veil of ignorance back and the frontier of discovery.” in front. And to be able to do that is like the professional honor of a lifetime.”
The concern here is not necessarily that OpenAI has developed superintelligence, which experts say is unlikely, but that the new breakthrough mentioned by Altman may have increased pressure on a company that is internally struggling to move forward safely (by its non-profit arm), but also make money (with its for-profit subsidiary). Altman also recently said that GPT-5, presumably a powerful successor to the alarming GPT-4, is currently in development.
As the news spread, as expected, some jokes were shared on social media. X user Shaurya wrote: “This is like the Roman Empire for people doing matrix multiplication.” And AI futurist Daniel Jeffries said: “The entire AI industry would like to thank the OpenAI board for giving us all the opportunity has to catch up.”
But not all reactions were doom and gloom. Over the course of Friday night, some at OpenAI made forward-looking statements. Evan Morikawa, engineering manager at OpenAI, wrote on research, product development and design. There is a clear internal consensus among these leaders that we are here for the larger mission.”
Expect to hear more from the OpenAI Board page as more details emerge.