The following was a presentation I made for Sören Elverlin's AI Safety Reading Group. I decided to draw everything by hand because powerpoint is boring. Thanks to Ben Pace for formatting it for LW! See also the IAF post detailing the research which this presentation is based on.
I don’t understand where that 1/2 comes from. Unless I have made a gross mistake P(A|A => B) < P(A) even if P(A&B) > P(A¬(B)). In your first example, if I swap P(AB) and P(A¬(B)) so that P(AB) = .5 and P(A¬(B))=.3 then P(A|A=>B) = .5/.7 ~ 0.71 < 0.8 = P(A).
This confused me as well. This being true ensures that the ratio P(A):P(not A) doubles at each step. But part of this comic seems to imply that being less than a half stops the trolling, when it should only stop the trolling from proceeding at such a fast-paced rate.
The following was a presentation I made for Sören Elverlin's AI Safety Reading Group. I decided to draw everything by hand because powerpoint is boring. Thanks to Ben Pace for formatting it for LW! See also the IAF post detailing the research which this presentation is based on.