My substantial crux on AI rights is I don't think you can actually make the AI lose their assets in the long run, and I also don't think you can make the AI economically interdependent once you've automated all jobs.
And I think the game-theoretic solutions fundamentally rest on assumptions about how much power actors have that I expect AI to break.
I've talked about why currently, property hasn't been expropriated from most humans, and inverting the points I made will make the case for why AI rights aren't reasonable (unless AIs are aligned beforehand):
My substantial crux on AI rights is I don't think you can actually make the AI lose their assets in the long run, and I also don't think you can make the AI economically interdependent once you've automated all jobs.
And I think the game-theoretic solutions fundamentally rest on assumptions about how much power actors have that I expect AI to break.
I've talked about why currently, property hasn't been expropriated from most humans, and inverting the points I made will make the case for why AI rights aren't reasonable (unless AIs are aligned beforehand):
https://www.lesswrong.com/posts/kgb58RL88YChkkBNf/?commentId=RQniGtFr2Fe4WCevR