Fundamentally, Sam Altman is a competent interpersonal operator. He’d doubtlessly worked both to be a naturally likeable and loyalty-inspiring person (in a passive way), and to purposefully inspire and select employees for loyalty (actively). That provided the backbone to the effort. No matter how many carrots and sticks were deployed, if Altman didn’t earn the loyalty to some extent, this show of support wouldn’t have been possible to achieve.
By contrast, the board apparently wasn’t very involved with the employees, and they did handle the communications terribly. Why would an OpenAI employee automatically assume they’re the good guys? (When even we are unsure.)
While loyalty to Altman might’ve varied, the employees for sure didn’t have any personal loyalty to the board.
And I’m sure there were carrots and sticks deployed aplenty:
On the carrots end: there could’ve been a ton of things, like Microsoft promising raises and guarantees if they jump ship (to make “we’d just quit otherwise” credible), Altman promising raises if he returns, arguments that they’d earn more in the long run if they either jump ship or get Altman back but not if they do nothing, etc.
Similarly, a ton of sticks: vivid images of the company imploding without Altman, and losing the investors, and of the board doing more random firings and wrecking things, plus covert suggestions of demotions or purges if people don’t support him and he comes back anyway, et cetera.
Which specific carrots and sticks were employed matters little, and likely differed from person to person to some extent. The point is just that there were a lot of things that could’ve sounded convincing with a right spin, and it was a relatively high-time-pressure situation, and the people spearheading the effort were (likely, apparently) good at making use of all of this.
The snowball effect/peer pressure obviously played a role. OpenAI employees are obviously different in how loyal/susceptible to the pressure they are, but as more and more people signed, the pressures would’ve mounted. First the 100 most loyal signed, then the 100 less-loyal ones (who wouldn’t have signed if the initiative didn’t already get some traction), then the 100 even-less-loyal ones, and so on.
If a random employee X were the first whom they asked to sign, that employee might or might not have refused. But if X is the seven-hundredth employee they’re asking, with 699 preceding ones having already signed, is X really going to make a stand? Yes, maybe! But that requires them to be against it on principle, such that they’re motivated to swim against the current.
And this effect could’ve been invoked even before the majority of the company signed – just by creating a narrative of the inevitability that of course we’re all gonna sign; that this is where the wind blows.
Lastly, I think signing the letter doesn’t actually commit an employee to anything. It’s not really legally binding? The cost of signing it is thus ~zero (except maybe some vague concerns about honor), whereas the supposed rewards for signing it and the punishments for not signing it (which, again, were doubtlessly floated around) are much more concrete.
And that point I’m outlining, in itself, is unlikely to be something the people organizing the effort had failed to empathize.
So there you have it: a relatively good boss is ousted by the board you know nothing about for unclear reasons, people close to the epicenter are running around telling you how it’s all going to implode now and how we have this costless way to maybe avert it, they’re being really pushy about it, it’s all very confusing and scary, more and more of the people around you are signing the letter, there’s an increasing atmosphere that signing it is just what an OpenAI employee does – would you really not sign?
Which isn’t to say it wasn’t an impressive accomplishment. The level of coordination required to pull this off was doubtlessly high, it would’ve required handling all of the aforementioned covert messaging about carrots-and-sticks with a minimal degree of competence, it required the foundation of Sam Altman establishing himself as a good leader, etc.
More or less all of it, I think.
Fundamentally, Sam Altman is a competent interpersonal operator. He’d doubtlessly worked both to be a naturally likeable and loyalty-inspiring person (in a passive way), and to purposefully inspire and select employees for loyalty (actively). That provided the backbone to the effort. No matter how many carrots and sticks were deployed, if Altman didn’t earn the loyalty to some extent, this show of support wouldn’t have been possible to achieve.
By contrast, the board apparently wasn’t very involved with the employees, and they did handle the communications terribly. Why would an OpenAI employee automatically assume they’re the good guys? (When even we are unsure.)
While loyalty to Altman might’ve varied, the employees for sure didn’t have any personal loyalty to the board.
And I’m sure there were carrots and sticks deployed aplenty:
On the carrots end: there could’ve been a ton of things, like Microsoft promising raises and guarantees if they jump ship (to make “we’d just quit otherwise” credible), Altman promising raises if he returns, arguments that they’d earn more in the long run if they either jump ship or get Altman back but not if they do nothing, etc.
Similarly, a ton of sticks: vivid images of the company imploding without Altman, and losing the investors, and of the board doing more random firings and wrecking things, plus covert suggestions of demotions or purges if people don’t support him and he comes back anyway, et cetera.
Which specific carrots and sticks were employed matters little, and likely differed from person to person to some extent. The point is just that there were a lot of things that could’ve sounded convincing with a right spin, and it was a relatively high-time-pressure situation, and the people spearheading the effort were (likely, apparently) good at making use of all of this.
The snowball effect/peer pressure obviously played a role. OpenAI employees are obviously different in how loyal/susceptible to the pressure they are, but as more and more people signed, the pressures would’ve mounted. First the 100 most loyal signed, then the 100 less-loyal ones (who wouldn’t have signed if the initiative didn’t already get some traction), then the 100 even-less-loyal ones, and so on.
If a random employee X were the first whom they asked to sign, that employee might or might not have refused. But if X is the seven-hundredth employee they’re asking, with 699 preceding ones having already signed, is X really going to make a stand? Yes, maybe! But that requires them to be against it on principle, such that they’re motivated to swim against the current.
And this effect could’ve been invoked even before the majority of the company signed – just by creating a narrative of the inevitability that of course we’re all gonna sign; that this is where the wind blows.
Lastly, I think signing the letter doesn’t actually commit an employee to anything. It’s not really legally binding? The cost of signing it is thus ~zero (except maybe some vague concerns about honor), whereas the supposed rewards for signing it and the punishments for not signing it (which, again, were doubtlessly floated around) are much more concrete.
And that point I’m outlining, in itself, is unlikely to be something the people organizing the effort had failed to empathize.
So there you have it: a relatively good boss is ousted by the board you know nothing about for unclear reasons, people close to the epicenter are running around telling you how it’s all going to implode now and how we have this costless way to maybe avert it, they’re being really pushy about it, it’s all very confusing and scary, more and more of the people around you are signing the letter, there’s an increasing atmosphere that signing it is just what an OpenAI employee does – would you really not sign?
Which isn’t to say it wasn’t an impressive accomplishment. The level of coordination required to pull this off was doubtlessly high, it would’ve required handling all of the aforementioned covert messaging about carrots-and-sticks with a minimal degree of competence, it required the foundation of Sam Altman establishing himself as a good leader, etc.
But I’m wholly unsurprised it worked.