fractalman comments on Prisoner's Dilemma (with visible source code) Tournament - Less Wrong

47 Post author: AlexMennen 07 June 2013 08:30AM

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

Comments (232)

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

Comment author: nshepperd 09 July 2013 12:55:08PM *  1 point [-]

Now that the contest is over, I will observe that contrary to your first claim...

MimicBots fit all of these parameters. They make it clear that their opponent is choosing between (C, C) and (D, D). Against a MimicBot, (C, D) is not an option.

...it's actually rather nontrivial to prove that your "MimicBot" does the same thing as you, since it doesn't run your program against itself, it runs your program against a different program. For example, PrudentBot defects against any of your MimicBots, since it can prove that "JusticeBot defects against me, since I defect against CooperateBot" therefore "JusticeBot+1 defects against me, since I defect against Justicebot", etc etc. In that sense, your mimicbot is not really a mimicbot at all. You could call it "simply a higher-order justicebot".

In contrast, with a mimicbot such as solipsist's one can just replace an instance of opponent(me) with 'C and see that this massively increases the chances of his bot cooperating, comparing to replacing the same thing with 'D, hence you should cooperate.

Comment author: fractalman 14 July 2013 06:21:48AM 0 points [-]

There's two types of mimicbots running around: fixed-rank, and random-reliant.
Which mimicbot are you analyzing?