Good article. I listened to all the rootclaim debate and found it informative. After that debate, I have a lot less belief in the credibility of giving accurate bayes estimates for complicated events, e.g. both debaters attempted it but their estimates where different by >>> like >1e20 I think.
I think this applies even more for P(doom) for AI, after all its about something that hasn’t even happened yet—I agree with the criticism that P(doom) is more a feeling rather than the result of rationality.
At one point Miller gave a likelihood against LL by a factor of 1e20 or 1e25, I think during the second debate, on genetic evidence. I don’t think he intended this number to be an actual Bayes factor, but rather to show how easy it is to get a big BF by multiplying many small numbers together (see also https://arbital.com/p/multiple_stage_fallacy/).
Is there a name for these sorts of errors of conjunction and disjunction in super high dimension parameter spaces? I usually just refer to it as ‘cold reading yourself.’
Good article. I listened to all the rootclaim debate and found it informative. After that debate, I have a lot less belief in the credibility of giving accurate bayes estimates for complicated events, e.g. both debaters attempted it but their estimates where different by >>> like >1e20 I think.
I think this applies even more for P(doom) for AI, after all its about something that hasn’t even happened yet—I agree with the criticism that P(doom) is more a feeling rather than the result of rationality.
At one point Miller gave a likelihood against LL by a factor of 1e20 or 1e25, I think during the second debate, on genetic evidence. I don’t think he intended this number to be an actual Bayes factor, but rather to show how easy it is to get a big BF by multiplying many small numbers together (see also https://arbital.com/p/multiple_stage_fallacy/).
Is there a name for these sorts of errors of conjunction and disjunction in super high dimension parameter spaces? I usually just refer to it as ‘cold reading yourself.’
https://arbital.com/p/multiple_stage_fallacy/ ?
Interesting that conjunctive fallacy is a broadly used term but disjunctive fallacy is not.
What would the disjunctive fallacy be? Failing to account for the fact that P(A or B) >= P(A) and P(B)?
assumption of independence