It looks like Sam Altman is using a bunch of “tricks” now trying to fight his way back into more influence over OpenAI. I’m not aware of anything I’d consider unethical (at least if one has good reasons to believe one has been unfairly attacked), but it’s still the sort of stuff that wouldn’t come naturally to a lot of people and wouldn’t feel fair to a lot of people (at least if there’s a strong possibility that the other side is acting in good faith too).
Many OpenAI employees have large monetary incentives on the line and there’s levels of peer pressure that are off the charts, so we really can’t read too much into who tweeted how many hearts or signed the letter or whatever.
Maybe the extent of this was obvious to most others, but for me, while I was aware that this was going on, I feel like I underestimated the extent of it. One thing that put things into a different light for me is this tweet.
Which makes me wonder, could things really have gone down a lot differently? Sure, smoking-gun-type evidence would’ve helped the board immensely. But is it their fault that they don’t have it? Not necessarily. If they had (1) time pressure (for one reason or another – hard to know at this point) and (2) if they still had enough ‘soft’ evidence to justify drastic actions. With (1) and (2) together, it could have made sense to risk intervening even without smoking-gun-type evidence.
(2) might be a crux for some people, but I believe that there are situations where it’s legitimate for a group of people to become convinced that someone else is untrustworthy without being in a position to easily and quickly convince others. NDAs in play could be one reason, but also just “the evidence is of the sort that ‘you had to be there’” or “you need all this other context and individual data points only become compelling if you also know about all these other data points that together help rule out innocuous/charitable interpretations about what happened.”
In any case, many people highlighted the short notice with which the board announced their decision and commented that this implies that the board acted in an outrageous way and seems inexperienced. However, having seen what Altman managed to mobilize in just a couple of days, it’s now obvious that, if you think he’s scheming and deceptive in a genuinely bad way (as opposed to “someone knows how to fight power struggles and is willing to fight them when he feels like he’s undeservedly under attack” – which isn’t by itself a bad thing), then you simply can’t give him a headstart.
So, while I still think the board made mistakes, I today feel a bit less confident that these mistakes were necessarily as big as I initially thought. I now think it’s possible – but far from certain – that we’re in a world where things are playing out the way they have mostly because it’s a really though situation for the board to be in even when they are right. And, sure, that would’ve been a reason to consider not starting this whole thing, but obviously that’s very costly as well, so, again, tough situation.
I guess a big crux is “how common is it that you justifiably think someone is bad but it’ll be hard to convince others?” My stance is that, if you’re right, you should eventually be able to convince others if the others are interested in the truth and you get a bunch of time and the opportunity to talk to more people who may have extra info. But you might not be able to succeed if you only have a few days and then you’re out if you don’t sound convincing enough.
My opinions have been fluctuating a crazy amount recently (I don’t think I’ve ever been in a situation where my opinions have gone up and down like this!), so, idk, I may update quite a bit in the other direction again tomorrow.
The board could (justifiably based on Sam’s incredible mobilization the past days**) believe that they have little to no chance of winning the war of public opinion and focus on doing everything privately since that is where they feel on equal footing.
This doesn’t explain fully why they haven’t stated reasons in private, but it does seem they provided at least something to Emmett Shear as he said he had a reason from the board that wasn’t safety or commercialization (PPS of https://twitter.com/eshear/status/1726526112019382275)
** Very few fires employees would even consider pushing back, but to be this successful this quickly is impressive. Not taking a side on it being good or evil, just stating the fact of his ability to fight back after things seemed gloom (betting markets were down below 10%)
Quite the saga: glad it is over and think that Larry Summers is a great independent thinker that could help the board make some smart expansion decisions
One thing I’ve realized more in the last 24h:
It looks like Sam Altman is using a bunch of “tricks” now trying to fight his way back into more influence over OpenAI. I’m not aware of anything I’d consider unethical (at least if one has good reasons to believe one has been unfairly attacked), but it’s still the sort of stuff that wouldn’t come naturally to a lot of people and wouldn’t feel fair to a lot of people (at least if there’s a strong possibility that the other side is acting in good faith too).
Many OpenAI employees have large monetary incentives on the line and there’s levels of peer pressure that are off the charts, so we really can’t read too much into who tweeted how many hearts or signed the letter or whatever.
Maybe the extent of this was obvious to most others, but for me, while I was aware that this was going on, I feel like I underestimated the extent of it. One thing that put things into a different light for me is this tweet.
Which makes me wonder, could things really have gone down a lot differently? Sure, smoking-gun-type evidence would’ve helped the board immensely. But is it their fault that they don’t have it? Not necessarily. If they had (1) time pressure (for one reason or another – hard to know at this point) and (2) if they still had enough ‘soft’ evidence to justify drastic actions. With (1) and (2) together, it could have made sense to risk intervening even without smoking-gun-type evidence.
(2) might be a crux for some people, but I believe that there are situations where it’s legitimate for a group of people to become convinced that someone else is untrustworthy without being in a position to easily and quickly convince others. NDAs in play could be one reason, but also just “the evidence is of the sort that ‘you had to be there’” or “you need all this other context and individual data points only become compelling if you also know about all these other data points that together help rule out innocuous/charitable interpretations about what happened.”
In any case, many people highlighted the short notice with which the board announced their decision and commented that this implies that the board acted in an outrageous way and seems inexperienced. However, having seen what Altman managed to mobilize in just a couple of days, it’s now obvious that, if you think he’s scheming and deceptive in a genuinely bad way (as opposed to “someone knows how to fight power struggles and is willing to fight them when he feels like he’s undeservedly under attack” – which isn’t by itself a bad thing), then you simply can’t give him a headstart.
So, while I still think the board made mistakes, I today feel a bit less confident that these mistakes were necessarily as big as I initially thought. I now think it’s possible – but far from certain – that we’re in a world where things are playing out the way they have mostly because it’s a really though situation for the board to be in even when they are right. And, sure, that would’ve been a reason to consider not starting this whole thing, but obviously that’s very costly as well, so, again, tough situation.
I guess a big crux is “how common is it that you justifiably think someone is bad but it’ll be hard to convince others?” My stance is that, if you’re right, you should eventually be able to convince others if the others are interested in the truth and you get a bunch of time and the opportunity to talk to more people who may have extra info. But you might not be able to succeed if you only have a few days and then you’re out if you don’t sound convincing enough.
My opinions have been fluctuating a crazy amount recently (I don’t think I’ve ever been in a situation where my opinions have gone up and down like this!), so, idk, I may update quite a bit in the other direction again tomorrow.
The board could (justifiably based on Sam’s incredible mobilization the past days**) believe that they have little to no chance of winning the war of public opinion and focus on doing everything privately since that is where they feel on equal footing.
This doesn’t explain fully why they haven’t stated reasons in private, but it does seem they provided at least something to Emmett Shear as he said he had a reason from the board that wasn’t safety or commercialization (PPS of https://twitter.com/eshear/status/1726526112019382275)
** Very few fires employees would even consider pushing back, but to be this successful this quickly is impressive. Not taking a side on it being good or evil, just stating the fact of his ability to fight back after things seemed gloom (betting markets were down below 10%)
Well, seems like the board did provide zero evidence in private, too! https://twitter.com/emilychangtv/status/1727228431396704557
Quite the saga: glad it is over and think that Larry Summers is a great independent thinker that could help the board make some smart expansion decisions
I feel like at this point the only truly rational comment is:
what the absolute fuck.
this Washington Post article supports the ‘Scheming Sam’ Hypothesis: anonymous reports mostly from his time at Y Combinator