This recent TIME article lays out the saga pretty straightforwardly and makes it a bit less confusing.
https://time.com/6342827/ceo-of-the-year-2023-sam-altman/
At least, that was how I felt after reading it.
Basically, within the span of a year, OpenAI transformed from a research lab inside a non-profit that was pursuing a seemingly-Quixiotic dream of artificial general intelligence (AGI)... into one of the fastest-growing for-profit software companies of all time via its creation of the chat-based generative AI category (aka ChatGPT) and its consumer/enterprise SaaS and API offerings.
The board -- or, at least, its 4 remaining non-CEO members -- thought that this was too much, too fast, and too soon, and that there was a narrow window of time where they could slow things down and refocus on the original non-profit mission. They also felt that Altman was a bit of a force of nature, had his own ideas about where OpenAI was going, and treated "board management" as one of his CEO skills to route around obstacles.
Once a board loses trust of their CEO, unfortunately, there is usually only one blunt and powerful tool left: firing the CEO.
And this happens pretty often. As the investor Jerry Neumann once put it, "Your board of directors is probably going to fire you."[1] Boards have very few ways to actually take action when they are worried about a company or institution; firing management is one of the few "course correction" actions they can take quickly.
In OpenAI's case, if they had a for-profit board, that board would probably have been ecstatic with Altman and the company's progress. But this was not a for-profit board. It was a non-profit (mission-oriented) board meant to oversee the safe rollout of AGI. Those board members weren't sure the best way to do that was to become one of the world's largest for-profit software companies.
I'd speculate that it was probably an emotional decision and the full implications were not entirely thought through until it was too late. I'd also speculate that this explains why Ilya Sutskever felt some immediate regret, because his goal wasn't to destroy OpenAI (or inspire an employee revolt) but to put its non-profit mission back into focus. I like to practice the principle of charity[2], and, in this case, I think the non-profit board was not acting maliciously but simply did not realize the knock-on effects of trying to replace a CEO when everything at the company seems to be "going right."
I suspect Altman thought the best way to roll out AI was via iterative product development and fast revenue growth to finance the GPU demands, utilizing corporate partnerships (Microsoft), viral word-of-mouth marketing, and SaaS/API fees (ChatGPT). Running out of data center compute started to become a primary concern, so it wouldn't surprise me if safety took a backseat to this. Remember, all this growth happened in the span of a year. Perhaps Altman thought he was satisfying the safety concerns simply by talking to regulators, making iterative releases, and going on a speaking tour about it, but the board thought the only way to go safely was to go slower. I'm sure we'll learn more after some books are written about the episode.
[1]: https://reactionwheel.net/2021/11/your-boards-of-directors-i...
[2]: https://en.wikipedia.org/wiki/Principle_of_charity