I did update a little bit against this when I heard (vaguely) that a lot of senior researchers at frontier AI labs do have kids. On the other hand, now that I think of it, I wonder if there’s kinda morally speaking two categories of AI capabilities researchers:
Older ones; tend to have kids, doesn’t worry about downside risks, doesn’t think AGI is a thing
Younger ones; tend to not have kids, kinda worries about downside risks, does think AGI is a thing, but has some rationalization because of Sinclair’s Razor or similar
I can imagine the first to be easier to switch to blocking AI progress once they feel the AGI than the latter as they don’t have rationalizations in place.
There’s also a third category, that of successionists which would ~likely change their minds if they had kids or if a lot of people dear to them had kids etc. as that (in my experience) moves your aliefs to “there’s something worth fighting for in humanity even if all current people are on average awful because the original state is so pure and full of potential for wonder and good things that a universe without this is one less valuable”
I can imagine the first to be easier to switch to blocking AI progress once they feel the AGI than the latter as they don’t have rationalizations in place.
That’s a good reason for hope! One would have thought they could be moved by the pedigree of those who have already flipped, such as Hinton etc., but here’s to hoping.
There’s also a third category, that of successionists which would ~likely change their minds if they had kids or if a lot of people dear to them had kids etc. as that (in my experience) moves your aliefs to “there’s something worth fighting for in humanity even if all current people are on average awful because the original state is so pure and full of potential for wonder and good things that a universe without this is one less valuable”
Plausible. Though to be honest, I have trouble getting into that mindset, and it’s plausible to me that it’s sufficiently skew to what seems natural to me, that it would not respond in this way. In other words, if someone is a staunch successionist, that suggests there’s already some wonky stuff generating that stance, which may not go away.
Are there many successionists at all? How common are they in actual AI capabilities research? Do they in fact tend to have kids? How old do they tend to be?
I did update a little bit against this when I heard (vaguely) that a lot of senior researchers at frontier AI labs do have kids. On the other hand, now that I think of it, I wonder if there’s kinda morally speaking two categories of AI capabilities researchers:
Older ones; tend to have kids, doesn’t worry about downside risks, doesn’t think AGI is a thing
Younger ones; tend to not have kids, kinda worries about downside risks, does think AGI is a thing, but has some rationalization because of Sinclair’s Razor or similar
I can imagine the first to be easier to switch to blocking AI progress once they feel the AGI than the latter as they don’t have rationalizations in place.
There’s also a third category, that of successionists which would ~likely change their minds if they had kids or if a lot of people dear to them had kids etc. as that (in my experience) moves your aliefs to “there’s something worth fighting for in humanity even if all current people are on average awful because the original state is so pure and full of potential for wonder and good things that a universe without this is one less valuable”
That’s a good reason for hope! One would have thought they could be moved by the pedigree of those who have already flipped, such as Hinton etc., but here’s to hoping.
Plausible. Though to be honest, I have trouble getting into that mindset, and it’s plausible to me that it’s sufficiently skew to what seems natural to me, that it would not respond in this way. In other words, if someone is a staunch successionist, that suggests there’s already some wonky stuff generating that stance, which may not go away.
Are there many successionists at all? How common are they in actual AI capabilities research? Do they in fact tend to have kids? How old do they tend to be?