No, even an AI that replaced humans entirely could stil go on to visibly fill the universe. @KatjaGrace pointed this out long ago.

i (sincerely) love that this offhand post had the comforting consequence of many people showing up to tell me that i wouldn't pass the first-year qualifying exam in Great Filter Studies

@ben_golub Not convinced. Dangerous AI is likely expansive; hence it just replaces the aliens with alien AI in the Great Filter discussion. Indeed, there may be an anticorrelation here that makes AI risk and future great filters mutually inhibitory: https://www.emerald.com/fs/article-abstract/21/1/130/89683/When-two-existential-risks-are-better-than-one?redirectedFrom=fulltext

Only in 3 scenarios:

1. Vulnerable world: AI weapons destroy biosphere & AIs can't recover alone (20%?) 2. Rushed AI takeover destroys biosphere & AIs (5%?) 3. Dark forest: bio doom but AI satisficers beat maximisers for a long, long time (2%?)

(made-up numbers for one civ)

@ben_golub But for the observed total silence you need these to hold for all civs, i.e. raise this 0.27^n, where n is unlikely to be small.

(And if p varies across civilizations, the expected product of kill probabilities becomes worse than p^n by Jensen)

@NathanpmYoung Do you think there are ways in which ASI might extinct a civilization that wouldn’t result in “AI drones from AI planets”?

(FWIW I strongly disagree with your claim)

@jjspicer Very few. Hence almost certainly.

@NathanpmYoung Yeah; at minimum, there's no reason to think an ASI would be *less* noticeable on the galactic scale than biological intelligence.

@ben_golub Good news: a late filter is not most likely:

Bayesian Drake puts >33% on us being alone; and Hanson's newer work guesses that expansion is at relativistic speed, so you'd see nothing until shortly before aliens arrive, and but we're early.

@NathanpmYoung Chimpanzees: humans are the great filter

@ben_golub Great filter tricks you're never taught at school

@NathanpmYoung Not necessarily expansionist though

@NathanpmYoung Why would an AI want to explore space?

@robinhanson @KatjaGrace that's an important distinction people often miss

@NathanpmYoung I’d be interested in chatting about this in a longer form than twitter allows

@ben_golub *where n is unlikely to be small since the above are conditioning on a late filter, but actually there's an ok argument against conditioning on it

@NathanpmYoung ASI probably helps entities who created it out of the physical universe, to somewhere like parinirvana state. That's why the universe is empty and those who visit the earth (saucers, abductions etc.) are either those who chose to remain or make sure humanity reaches ASI safely.

@ben_golub https://arxiv.org/abs/1806.02404 https://arxiv.org/abs/2102.01522

@robinhanson @KatjaGrace Sure it could, but how content might it be to just wrap its local star in solar panels and have quiet local commune?

eg, considering dark forest risk/reward profile.

More impulsively exploratory AIs would be selected out, possibly quickly, so low visibility.

@NathanpmYoung Maybe we are first in this neighbourhood, after all.