Cited Excerpts from The Optimist by Keach Hagey
"One night in the summer of 2023, an OpenAI board member overheard a person at a dinner party discussing how inappropriate it was that returns from OpenAI’s Startup Fund were not going to OpenAI investors, given how the fund was able to use scarce resources such as early access to OpenAI products. This was also news to the board. What was this person talking about? OpenAI did launch a startup fund in 2021, which it said at the time would be 'managed' by OpenAI, with investment from Microsoft and others. The fund had invested in the AI-driven legal startup Harvey and a handful of other companies. But what was this about returns not going to OpenAI shareholders? The board began to ask Altman about it, and over the course of months of going back and forth, eventually learned that Altman owned the fund personally, and had been raising money from LPs to fund it … OpenAI has said Altman had no financial stake in the fund, and had just set it up personally because it was the fastest way to do it. (Initial answers to the board had been about the tax advantages of this structure.) But that would make a weirdly structured fund even weirder. The independent board members felt like they could not get a straight answer. They also felt they should have been informed ahead of time, given Altman’s repeated claim to not have a stake in OpenAI—and how crucial this status was to his ability to serve on the board at all."
“It had taken Sutskever years to be able to put his finger on Altman’s pattern of behavior—how OpenAI’s CEO would tell him one thing, then say another and act as if the difference was an accident. ‘Oh, I must have misspoken,’ Altman would say. Sutskever felt that Altman was dishonest and causing chaos, which would be a problem for any CEO, but especially for one in charge of such potentially civilization-altering technology.”
“Murati described how what she saw as Altman’s toxic style of running the company had been causing problems for years, especially when his anxiety flared up, such as in recent months. In her experience, Altman had a simple playbook: first, say whatever he needed to say to get you to do what he wanted, and second, if that didn’t work, undermine you or destroy your credibility. Altman told people what they wanted to hear, even if it meant promising two competing people the same job or giving Microsoft some ground on a negotiation that she had spent months trying to gain. She explained how the dynamic between Brockman and Altman made it almost impossible for her to do her job: Brockman, who was the president but had no direct reports, technically reported to her, yet was also on the board. Whenever she tried to keep him from bigfooting someone else’s project, he just went to Altman and got around her. Brockman’s behavior had pushed multiple people out of the company, and the unsustainable dynamics between Altman and Brockman were about to push out several more, including chief research officer Bob McGrew and maybe even Murati herself. Murati had given Altman feedback on all these points in the past, she said. Altman had responded by bringing the company’s head of HR, Diane Yoon, to their one-on-ones for weeks until she finally told him that she didn’t intend to share her feedback with the board.”
“In some instances, Microsoft fell afoul of the DSB, but the OpenAI board was alarmed when they were informed about such a setback from an employee—who stopped a board member in the hallway and asked if the board knew about the safety breach—rather than from Altman, despite having just completed a six-hour board meeting. In late 2022, Microsoft had rolled out a version of still-unreleased GPT-4 in a test in India without getting DSB approval first. While it ultimately got it, the breach in India suggested to some board members that the companies’ safety processes were not working.”
“In April 2023, Altman wrote his fellow board members to say that D’Angelo’s involvement with Poe had become a true conflict, and it was time for him to leave the board. Didn’t they agree? Toner and McCauley did not. The board had just spent weeks discussing what constituted a conflict and decided that it had to be a competitor training a rival frontier model, meaning one at the cutting edge of AI science. Since Poe was just a wrapper, it seemed to fall short of the bar. Brockman then chimed in with a different reason why D’Angelo should go: Poe was a customer, and thus it was a conflict for D’Angelo to have information about OpenAI’s internal business.”
“The nonprofit structure proved to be a helpful selling point in recruiting AI researchers. ‘The pitch was, ‘AI is going to be so important. We don’t want a for-profit entity to be in charge of this, because of the profit-seeking motives that are just never-ending,’ Karpathy said. ‘That appealed to me quite a bit.’”
“It was only because the board members were having so much regular contact with each other that they caught Altman in the lie that ended up being the nail in the proverbial coffin. During Sutskever’s conversation with McCauley in late October, he mentioned that Altman had told him that McCauley had told Altman that Toner should obviously leave the board over the article. McCauley knew she had said no such thing. She called D’Angelo and explained what had happened. ‘That’s very concerning,’ he said. In the minds of the company’s three independent board members, the incident seemed to crystalize Altman’s MO of placing his true wishes in the mouths of others in order to continue to be liked by everybody.”
“To prove his point, Sutskever emailed Toner, McCauley, and D’Angelo two lengthy documents using Gmail’s self-destructing pdf function. One was about Altman, the other about Brockman. The Altman document consisted of dozens of examples of lying or other toxic behavior, largely backed up by screenshots from Murati’s Slack channel that Sutskever had compiled. In one of them, Altman tells Murati that the company’s legal department had said that GPT-4 Turbo didn’t need to go through DSB review. When Murati checks with Jason Kwan, the company’s top lawyer, he is confused, and says he can’t imagine how Altman would have gotten that impression; of course GPT-4 Turbo had to go through DSB.”
