I don’t know if and where in OpenAI’s 3180, 18th Street, San Francisco office building Sam Altman is liable to bump into Ilya Sutskever again. But that bagel’n’latte-armed encounter will be awkward. Very awkward.
On Monday, Sutskever, who along with Altman and Greg Brockman founded AI’s version of the Manhattan Project, had strangely added his name to a letter with more than 700 OpenAI employees, demanding that the board of directors that sacked Altman two days before, resign. This was strange because two days before the letter stating that the board was ‘incapable of overseeing OpenAI’ was sent off, Sutskever was one of the board’s directors who had lost confidence in Altman.
On X (formerly Twitter), Sutskever now Xed (formerly, tweeted) ‘I deeply regret my participation in the board’s actions. I never intended to harm OpenAI. I love everything we’ve built together and I will do everything I can to reunite the company.’ With Satya Nadella having provided Altman at that point a garage within Microsoft to work his AI experiments from, did Sutskever find himself on the road to Damascus, and throw thirty pieces of silver back at the board and become a born-again Altmanirbhar?
But even as he finally re-read the tea leaves right about Altman returning from (a 4-day) vanvas as the righteous king of OpenAIyodhya, Altman 2.0 sacked the whole board by Thursday (barring Quora CEO Adam D’Angelo), leaving us with the updated motto: There’s a Sutskever born every minute.
But cut through the fog of the OpenAI boardroom, and the smog still remains: why did the board lose confidence in Altman in the first place? Altman, according to the board’s statement, ‘was not consistently candid in his communications with the board, hindering its ability to exercise its responsibilities’. Who on Earth is ever candid on a board? That would be like expecting Salman Khan to play the role of the Dark Knight to Christopher Nolan‘s satisfaction. (Not a bad idea, actually, Anurag?)The board is the last place one is candid. Its main function is to provide its members a table for virtue signalling. This is where matters like ESG, reducing the gender pay divide, bringing peace to Gaza, providing the first-floor staff with better coffee is paid lip gloss service, while everyone speaks ‘candidly’ only in ayes and nays when voting comes. For the board to hang Altman for not ‘speaking his mind’ is like Al Capone being jailed not even for tax evasion, as he was, but for shoplifting. Much speculation is being made about whether the board thought Altman was conducting the equivalent of ‘Nazi experiments’ on the side, putting humanity — the directors saw themselves representing humanity like some dinky edition of Star Trek – in existential danger. They may have thought that AI’s poster boy was too much in a hurry, looking at putting out too many products like ChatGPT and DALL-E out on the shelves without taking precautions against an AI ‘entity’ that will become self-aware by the time Trump returns and trigger a nuclear war to exterminate the human species.
Altman himself has publicly warned against this possibility publicly, including at a conversational event with ET earlier this year. But the board may have known something about Altman being Dr Strangelove that he wasn’t candid about. Again, strangely, the now-destroyed OpenAI board was not candid about what they meant about Altman not being ‘consistently candid’. Was he candid in fits and starts, spluttering, ‘I hav bin creating monzters!’ or ‘Mein Gott, ve need to make a lot of money to ensure zat ve reach our goal of creating artificial general intelligence [a program that outperforms humans on intellectual tasks]!’ in between, ‘Gentlemen, ladies, laterz.’
In Voltaire’s great 1759 satire, Candide: Or, Optimism, the eponymous hero is disillusioned with his vision of an Edenic paradise by the harsh realities of the world. Candide’s valet, Cacambo, at one point asks him, ‘Optimism, what is that?’ ‘Alas!’ replied Candide, ‘It is the obstinacy of maintaining that everything is best when it is worst.’ Perhaps, this is what Altman was not consistently candid about when hanging with the ex-OpenAI board.
