[chr1]

> I'm not sure what nuclear deterrent looks like when you have AGI resources.

Other people who have AGI resources. That's why it is very important to have multiple groups of people who will reach AGI at the same time. That way the same iterated prisoners dilemma thing will act on AGIs too.

[moonchrome]

I'm not sure how that doesn't devolve into an AGI armageddon. Those self improving AGIs would have to constantly be on equal level to avoid takeover, and they have an existential imperative to one up each other.

There's no cooperation incentive between them just a potential stalemate.

Also the stalemate covers the people in control of AGIs (assuming they aren't acting independently) - the rest of us are just resources in breaking the stalemate game.

This sounds like the best way to build paperclip maximizer.

[boringuser2]

It's pretty clear that at some point in the process of recursive self-improvement that you'd have literally just invented God, whereafter all predictions break down.

[mrob]

The more groups have AGI, the higher the chance that one of them figures out recursive self-improvement, and if that happens it's game over for everybody. Iterated prisoner's dilemma only works when all the prisoners have roughly the same power.

[chr1]

Why is it game over though? AI is already used for self-improvement, and will be used more as it becomes more capable, but that won't magically solve all the problems in the world and create omnipotent AI. If there are many people and many AI's they collectively will be stronger than the one most advanced, and cooperation will remain advantageous.