“What We Know So Far About Why OpenAI Fired Sam Altman”, Max Chafkin, Rachel Metz2023-11-19 (OA)⁠:

…And yet on November 6, at the company’s first developer conference, the acclaim for Sam Altman seemed universal. Attendees applauded rapturously as he ticked off the company’s accomplishments: 2 million customers, including “over 92% of Fortune 500 companies.” A big reason for that was Microsoft which invested $13 billion into the company and put Altman at the center of a corporate overhaul that has caused it to leapfrog rivals like Google and Amazon in certain categories of cloud computing, reinvigorated its Bing search engine, and put the company in the leading position in the hottest software category. Now, Altman invited CEO Satya Nadella onto the stage and asked him how Microsoft felt about the partnership. Nadella started to respond, and then broke into laughter, as if the answer to the question was absurdly obvious. “We love you guys”, he finally said after he’d calmed down. He thanked Altman for “building something magical.”

But if customers and investors were happy, there was one constituency that remained deeply skeptical of Altman and the very idea of a commercial AI company: Altman’s own board of directors. Although the board included Altman and a close ally, OpenAI President Greg Brockman, it was ultimately controlled by the interests of scientists who worried that the company’s expansion was out of control, maybe even dangerous.

That put the scientists at odds with Altman and Brockman, who both argued that OpenAI was growing its business out of necessity. Every time a customer asks OpenAI’s ChatGPT chatbot a question it requires huge amounts of expensive computing power—so much that the company was having trouble keeping up with the explosive demand from users. The company has been forced to place limits on the number of times users can query its most powerful AI models in a day. In fact, the situation got so dire in the days after the developer conference, Altman announced that the company was pausing sign-ups for its paid ChatGPT Plus service for an indeterminate amount of time.

From Altman’s point of view, raising more money and finding additional revenue sources were essential.

…The board had also moved without consulting with Microsoft, leaving Nadella “livid” at the hasty termination of a crucial business partner, according to a person familiar with his thinking. Nadella was “blindsided” by the news, this person said.

According to people familiar with his plans, Altman was plotting a competing company, while investors were agitating for his restoration.

Over the weekend, some investors were considering writing down the value of their OpenAI holdings to zero, according to a person familiar with the discussions. The potential move, which would both make it more difficult for the company to raise additional funds and allow OpenAI investors to back Altman’s theoretical competitor, seemed designed to pressure the board to resign and bring Altman back.

[VC funds typically have policies against investing in direct competitors due to conflicts of interest. So a VC fund like Founders fund which invested in OA could not invest in, say, Anthropic: they are “conflicted out”. But abandoning their equity frees them to invest in a competitor.]

Meanwhile, on Saturday night, numerous OpenAI executives and dozens of employees started tweeting the heart emoji—a statement of solidarity that appeared equal parts an expression of love for Altman and a rebuke to the board.

A source familiar with Nadella’s thinking said that the Microsoft CEO was advocating for Altman’s potential return and would also be interested in backing Altman’s new venture. The source predicted that if the board doesn’t reconsider, a large continent of OpenAI engineers would likely resign in the company days. Adding to the sense of uncertainty: OpenAI’s offices are closed all this week. Microsoft and Altman declined to comment. When reached by phone on Saturday, Brockman, who resigned shortly after Altman was fired, said “Super heads down right now, sorry.” Then he hung up.

A philosophical disagreement wouldn’t normally doom a company that had been in talks to sell shares to investors at an $86 billion valuation, but OpenAI was nothing like a normal company. Altman structured it as a nonprofit, with a for-profit subsidiary that he ran and that had aggressively courted venture capitalists and corporate partners. The novel—and, as OpenAI critics see it, flawed—structure put Altman, Microsoft, and all of the company’s customers at the mercy of a wonky board of directors that was dominated by those who were skeptical of the corporate expansion.

…Despite his position as founder and CEO, Altman has said he holds no equity in the company, framing this as of a piece with the company’s philanthropic mission. [Holding no equity permits him to hold his own board seat.] But of course, this would-be philanthropy had also sold 49% of its equity to Microsoft, which was granted no seats on its board. In an interview earlier this year, Altman suggested that the only recourse Microsoft had to control the company would be to unplug the servers that OpenAI rented. “I believe they will honor their contract”, he said at the time.

…The ultimate power at the company rested with the board, which included Altman, Ilya Sutskever and Brockman. The other members were Quora CEO Adam D’Angelo, tech entrepreneur Tasha McCauley and Helen Toner, director of strategy at Georgetown’s Center for Security and Emerging Technology. McCauley and Toner both had ties to effective altruism nonprofits. Toner had previously worked for Open Philanthropy; McCauley serves on the boards of Effective Ventures and 80,000 Hours.

…At the same time Altman was pursuing side projects that had the potential to enrich him and his investors, but which were outside of the control of OpenAI’s safety-conscious board. There was Worldcoin, his eyeball-scanning crypto project, which launched in July and was promoted as a potential universal basic income system to make up for AI-related job losses. Altman also explored starting his own AI chipmaker [to replace Nvidia GPUs], pitching sovereign wealth funds in the Middle East on an investment that could reach into the tens of billions of dollars, according to a person familiar with the plan. He also pitched SoftBank Group led by Japanese billionaire and tech investor Masayoshi Son, on a potential multibillion-dollar investment in a company he planned to start with former Apple design guru Jony Ive to make AI-oriented hardware.

…These efforts, along with the for-profit’s growing success, put Altman at odds with Sutskever, who was becoming more vocal about safety concerns. In July, Sutskever formed a new [Superalignment] team within the company focused on reining in “super intelligent” AI systems of the future. Tensions with Altman intensified in October, when, according to a source familiar with the relationship, Altman moved to reduce Sutskever’s role at the company [by promoting Jakub Pachocki], which rubbed Sutskever the wrong way and spilled over into tension with the company’s board.

At the event on Nov. 6, Altman made a number of announcements that infuriated Sutskever and people sympathetic to his point of view, the source said. Among them: customized versions of ChatGPT, allowing anyone to create chatbots that would perform specialized tasks. OpenAI has said that it would eventually allow these custom GPTs to operate on their own once a user creates them. Similar autonomous agents are offered by competing companies but are a red flag for safety advocates.

…In the days that followed, Sutskever brought his concerns to the board. According to an account posted on Twitter by Brockman, Sutskever texted Altman the evening of Nov. 16, inviting him to join a video call with the board the following day. Brockman was not invited. The following day at noon, Altman appeared and was told he was being fired. Minutes later, the announcement went out and chaos followed.

…If Altman does get his job back, Musk said he’s “very worried”, he posted on Twitter on Sunday. “Ilya has a good moral compass and does not seek power. He would not take such drastic action unless he felt it was absolutely necessary.” [cf. Ilya’s reasons for reversal]