w/r to zero not being a probability: obviously. The probability is extremely low, not zero, such that the chance of a benevolent AI existing is greater than the chance of humanity surviving a single malevolent AI. If that's not the case, then 1 and 2 are useless.
Peter, thanks. After reading that Drexler piece, the linked Christiano piece, the linked Eliezer post, and a few others, especially the conversation between Eliezer and Drexler in the comments of Drexler's post, I agree with you. TBH I am surprised that there's no better standard argument in support of inevitable anti-human collusion than this from Eliezer: "They [AIs] cooperate with each other but not you because they can do a spread of possibilities on each other modeling probable internal thought processes of each other; and you can’t adequately well-model a spread of possibilities on them, which is a requirement on being able to join an LDT coalition." As Christiano says, that makes a lot of assumptions.
Thanks for the replies.
w/r to zero not being a probability: obviously. The probability is extremely low, not zero, such that the chance of a benevolent AI existing is greater than the chance of humanity surviving a single malevolent AI. If that's not the case, then 1 and 2 are useless.
Peter, thanks. After reading that Drexler piece, the linked Christiano piece, the linked Eliezer post, and a few others, especially the conversation between Eliezer and Drexler in the comments of Drexler's post, I agree with you. TBH I am surprised that there's no better standard argument in support of inevitable anti-human collusion than this from Eliezer: "They [AIs] cooperate with each other but not you because they can do a spread of possibilities on each other modeling probable internal thought processes of each other; and you can’t adequately well-model a spread of possibilities on them, which is a requirement on being able to join an LDT coalition." As Christiano says, that makes a lot of assumptions.