Some differences: TDT’s “self” category cares about different things (something with logical dependence may still be too much of a departure from you to be special to your values); and if you think it possible for a mind without morally relevant experience to exist then the “other” vs “dumb stuff” category is also different. However, the similarity does still seem hopeful; perhaps humanity “lucked out” and got easier to formalize values than most possible minds. Our values are the way they are because of how evolution played out on Earth; while decision theory doesn’t recover morality, it looks like the three categories are there for basically the same reasons. This suggests that even if aliens have substantial value differences from us, they might also have a similar lucky split. Or maybe not, who knows.
Extra edit note: It looks like infrabayesian physicalism only (presently) deals with ‘computationalist’ values. This is stronger evidence for alignment being easier for humans than the glucose-maximizing-aliens (who only care about sugar in the ‘top layer’ of reality, and wouldn’t care about an atomic simulation of sugar).
Some differences: TDT’s “self” category cares about different things (something with logical dependence may still be too much of a departure from you to be special to your values); and if you think it possible for a mind without morally relevant experience to exist then the “other” vs “dumb stuff” category is also different. However, the similarity does still seem hopeful; perhaps humanity “lucked out” and got easier to formalize values than most possible minds. Our values are the way they are because of how evolution played out on Earth; while decision theory doesn’t recover morality, it looks like the three categories are there for basically the same reasons. This suggests that even if aliens have substantial value differences from us, they might also have a similar lucky split. Or maybe not, who knows.
Extra edit note: It looks like infrabayesian physicalism only (presently) deals with ‘computationalist’ values. This is stronger evidence for alignment being easier for humans than the glucose-maximizing-aliens (who only care about sugar in the ‘top layer’ of reality, and wouldn’t care about an atomic simulation of sugar).
Aren’t I glad I don’t have a sweet tooth?