Hm… this isn’t a dig at you or anything, just a thought I’m having trouble answering for myself right now, but how low does one get to set the probability of a prediction, and still call it a prediction? At which point—seems it’s not 50% - should you rather say “I’m predicting this won’t happen?”
If your prediction is lower than 50%, what you’re really saying is, “Of all the hypotheses I that have been elevated to my attention, this one is most likely; however, I am so uncertain that I am more likely to be wrong than right.”
Or in other words, to paraphrase Eliezer, I’m fairly sure that random person’s name isn’t Klein, but I’m very sure it’s not Ktlzybplq.
Imagine that Eliezer has 100 possible ways to complete this puzzle. 25% means that the combined weight of all the other possibilities outweighs the one I’ve outlined, but I am still privileging the one significantly over a typical possibility. (However, lerjj has convinced me that I was most likely wrong to do so.)
It’s useful to be able to simply focus on probability values it makes things less complicated than always having to shift when the probability goes beyond 50%.
Hm… this isn’t a dig at you or anything, just a thought I’m having trouble answering for myself right now, but how low does one get to set the probability of a prediction, and still call it a prediction? At which point—seems it’s not 50% - should you rather say “I’m predicting this won’t happen?”
If your prediction is lower than 50%, what you’re really saying is, “Of all the hypotheses I that have been elevated to my attention, this one is most likely; however, I am so uncertain that I am more likely to be wrong than right.”
Or in other words, to paraphrase Eliezer, I’m fairly sure that random person’s name isn’t Klein, but I’m very sure it’s not Ktlzybplq.
I’ve spent too long on LW, I think—I saw “Ktlzybplq” and assumed it was rot13.
It’s almost Mxyzptlk.
Heh, had the same thing. I already figured it wouldn’t be rot13, but I was still uncertain enough that I went and checked.
Imagine that Eliezer has 100 possible ways to complete this puzzle. 25% means that the combined weight of all the other possibilities outweighs the one I’ve outlined, but I am still privileging the one significantly over a typical possibility. (However, lerjj has convinced me that I was most likely wrong to do so.)
Right. What I was thinking, except for some reason I had a complete block on putting that thought into words. Thanks.
It seems like it would be determined by the relevant maximum entropy distribution.
It’s useful to be able to simply focus on probability values it makes things less complicated than always having to shift when the probability goes beyond 50%.