Eliezer_Yudkowsky comments on Expected Creative Surprises - Less Wrong

25 Post author: Eliezer_Yudkowsky 24 October 2008 10:22PM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (41)

Sort By: Old

You are viewing a single comment's thread.

Comment author: Eliezer_Yudkowsky 25 October 2008 02:19:37PM 1 point [-]

If you use upper-lower then you will forego combinations of bets that are sure gains. Dutch Book is inescapable, and only your presumption that the environment will contain agents who try to offer you losing bet combinations but no agents who are nice enough to take them, would lead you think that upper-lower "solves" the problem. No real agent can afford the computing power to be an ideal Bayesian, but the laws are ineluctable: whoever is not Bayesian - regardless of the reasons, regardless of whether it is unavoidable - is losing something.