How do the AGI’s model each other accurately enough to be able to acausally trade with each other like that? Is just using UDT/TDT enough? Probably. Is every sufficiently intelligent AGI going to switch to that, regardless of the decision theory it started out with, the way a CDT AGI would? Maybe there are possible alien decision theories that don’t converge that way but are still winning enough to be a plausible threat?
How do the AGI’s model each other accurately enough to be able to acausally trade with each other like that? Is just using UDT/TDT enough? Probably. Is every sufficiently intelligent AGI going to switch to that, regardless of the decision theory it started out with, the way a CDT AGI would? Maybe there are possible alien decision theories that don’t converge that way but are still winning enough to be a plausible threat?