orthonormal comments on Decision Theories, Part 3.75: Hang On, I Think This Works After All - Less Wrong

23 Post author: orthonormal 06 September 2012 04:23PM

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

Comments (45)

You are viewing a single comment's thread. Show more comments above.

Comment author: orthonormal 15 September 2012 09:46:39PM 0 points [-]

Thanks for the reply, and sorry for the delay...

Masquerade correctly defects against TemptingCooperateBot, for the following reasons:

  • PA proves "FB(TCB)=TCB(FB)=C"
  • PA+5 proves "TCB(DB)=C" and of course PA proves "DB(TCB)=D"
  • PA+5 proves "TCB(Masquerade)=C" (though this is actually irrelevant, as Masquerade will defect at this point whether the sanity check succeeds or fails)

It's not that there's anything magical about PA+4 in Masquerade; my mental image is that Masquerade uses a large value for N, so that it has time to find mutual cooperation with all but the very loopiest of opponents. If you delay it until PA+4, then it'll just start from there, more or less.