You have just been Counterfactually Mugged!
post by CronoDAS · 2009-08-19T22:24:38.805Z · LW · GW · Legacy · 25 commentsContents
25 comments
I'm going to test just how much the people here are committed to paying a Counterfactual Mugger, by playing Omega.
I'm going to roll a die. If it doesn't come up 5 or 6, I'm going to ask Eliezer Yudkowsky to reply to this article with the comment "I am a poopy head." If I roll a 5 or 6, I'm going to donate $20 to SIAI if I predict that Eliezer Yudkowsky will post the above comment.
Because Eliezer has indicated that he would pay up when counterfactually mugged, I do predict that, if I roll a 5 or 6, he'll respond.
::rolls die::
Darn it! It's a 5. Well, I'm a man of my word, so...
::donates::
Um, let's try that again. (At least I've proven my honesty!)
::rolls die::
Okay, this time it's a 1.
So, Eliezer, will you post a comment admitting that you're a poopy head?
25 comments
Comments sorted by top scores.
comment by Eliezer Yudkowsky (Eliezer_Yudkowsky) · 2009-08-19T23:08:16.922Z · LW(p) · GW(p)
I will if Michael Vassar judges that any reputational damage from the comment has an expected value less than $14.
You did it wrong on two counts: First, you need to ask me to pay you money, so the two utilities are easily commensurable and there's no question of interpreting the results. Second, repeating the Counterfactual Mugging more than once tends to obscure the point, especially given the implication that you had a stopping algorithm rather than a fixed number of iterations. Of course it is now too late to do it over again correctly.
But with a trusted witness of the original die roll, or say paying $20 if the 100th decimal digit of pi (unknown to me currently) is 0, and otherwise demanding $1, we could totally mug, say, Derek Parfit and see what happens. Actually, I think I'll forward this suggestion to Anders Sandberg and see what happens if he mugs Nick Bostrom. No one tell Bostrom before then, please.
Replies from: Eliezer_Yudkowsky, Peter_de_Blanc, CronoDAS, Larks, JamesAndrix, SforSingularity↑ comment by Eliezer Yudkowsky (Eliezer_Yudkowsky) · 2014-03-09T18:55:27.495Z · LW(p) · GW(p)
Historical note: We ran this, the 100th decimal digit of pi was not 0, and Bostrom paid Sandberg, which Sandberg had correctly predicted Bostrom would do.
↑ comment by Peter_de_Blanc · 2009-08-20T01:00:58.737Z · LW(p) · GW(p)
I will if Michael Vassar judges that any reputational damage from the comment has an expected value less than $14.
Don't you mean $10?
Replies from: Eliezer_Yudkowsky↑ comment by Eliezer Yudkowsky (Eliezer_Yudkowsky) · 2009-08-20T05:40:35.065Z · LW(p) · GW(p)
2/3 of $20? Should be "less than $13", actually.
Replies from: Peter_de_Blanc, Benquo↑ comment by Peter_de_Blanc · 2009-08-20T11:48:05.643Z · LW(p) · GW(p)
Before the die roll, there's a 1/3 chance that you'll get the reward ($20), and a 2/3 chance that you'll be asked to pay the penalty. For the expected utility to be 0, (2/3)*|penalty| = (1/3)*$20. Multiply both sides by 3, and 2*|penalty| = $20, so |penalty| = $10.
Replies from: Benquo↑ comment by Benquo · 2009-08-20T20:29:40.364Z · LW(p) · GW(p)
How's that for minor reputational damage?
Replies from: Eliezer_Yudkowsky↑ comment by Eliezer Yudkowsky (Eliezer_Yudkowsky) · 2009-08-20T21:58:50.589Z · LW(p) · GW(p)
Crap. Never mind. You know, this happens when I'm sufficiently tired, and it's scary. I am a poopy head.
Replies from: CronoDAS↑ comment by CronoDAS · 2009-08-20T18:06:17.508Z · LW(p) · GW(p)
But if I had stopped at just one roll of 5, I wouldn't have been able to make this post, and that wouldn't have been any fun!
Anyway, I thought that making a silly comment would be less painful than giving up actual money. It's not like we've been trying hard to present ourselves in a particularly dignified manner on this blog, anyway. What's wrong with wearing the clown suit once in a while?
Replies from: John_Maxwell_IV↑ comment by John_Maxwell (John_Maxwell_IV) · 2009-08-20T21:49:57.158Z · LW(p) · GW(p)
It's not like we've been trying hard to present ourselves in a particularly dignified manner on this blog, anyway.
I place importance on the opinions of outsiders who stumble across the site and see the things I say associated with my real name. Maybe we need an invite-only Lesswrong Premium™ where we sit around in clown suits discussing our sex lives (or lack thereof).
↑ comment by Larks · 2009-08-20T23:06:12.276Z · LW(p) · GW(p)
Second, repeating the Counterfactual Mugging more than once tends to obscure the point,
Surely if you'd take the bet once, you'd take it any number of times, so this shouldn't make much of a difference- and I can't see how stopping algorithm / fixed number of rolls makes a difference.
↑ comment by JamesAndrix · 2009-08-20T15:55:34.581Z · LW(p) · GW(p)
I would guess more reputational damage if you don't.
There's got to be a way to get positive press about a rash of muggings by intellectuals against other intellectuals.
↑ comment by SforSingularity · 2009-08-19T23:15:45.131Z · LW(p) · GW(p)
Expected reputational damage is probably $a few hundred IMO. It could be picked up and used as ammunition against SIAI.
Replies from: JamesAndrix↑ comment by JamesAndrix · 2009-08-20T16:11:48.966Z · LW(p) · GW(p)
Few will care about a silly comment on one post, especially given the context. Anybody using it as ammunition will look really petty.
Better ammunition:"But when Eliezer was counterfactually mugged, he didn't pay. How serious is he on his decision theory?"
Even if the procedural problems here give him a good reason not to 'pay', paying is better rhetorically. (though he may clarify that he won't pay under similarly sloppy conditions again.)
Replies from: Eliezer_Yudkowsky, John_Maxwell_IV, SforSingularity↑ comment by Eliezer Yudkowsky (Eliezer_Yudkowsky) · 2009-08-20T18:25:29.702Z · LW(p) · GW(p)
Even if the procedural problems here give him a good reason not to 'pay', paying is better rhetorically.
But that isn't supposed to be part of the Counterfactual Mugging scenario! Anyone would pay then.
↑ comment by John_Maxwell (John_Maxwell_IV) · 2009-08-20T21:52:25.667Z · LW(p) · GW(p)
Few will care about a silly comment on one post, especially given the context. Anybody using it as ammunition will look really petty.
I disagree. I think it could be used very effectively without explanation alongside a more serious criticism, possibly with a permalink to the actual comment.
↑ comment by SforSingularity · 2009-08-20T20:42:56.194Z · LW(p) · GW(p)
it would all be more clear-cut if the amounts involved were a few hundred $. But yes, good point.
comment by eirenicon · 2009-08-20T00:10:13.220Z · LW(p) · GW(p)
Hello, CronoDAS. I am Omegeirenicon. Since I am a powerful and prescient space brain, I know you are an infamous karma whore, so I propose a game. I will flip a coin, and if it comes up heads, I will vote up every single article and comment you have ever made. If it comes up tails, you will vote up this comment. I normally do not play games*, but I have predicted you will agree to these terms. Also, I already flipped the coin, and it came up tails. Thanks for your time.
*Yeah, right.
Replies from: CronoDAS, Cyancomment by Christian_Szegedy · 2009-08-25T21:47:21.587Z · LW(p) · GW(p)
I seems more like a factual bribing than a counterfactual mugging. :)
comment by cousin_it · 2009-08-22T21:55:38.768Z · LW(p) · GW(p)
How to do it properly:
Choose a list of victims in advance, and mug each of them once and only once - if and only if you honestly feel that they would pay up if asked. (If you feel they wouldn't, it might be amusing to send them an email describing the chance they missed.) Don't publish the list; each individual mugging must come as a surprise to the muggee. Afterwards the statistics of die rolls will prove or disprove your honesty, and the statistics of responses will prove or disprove the honesty of victims (or maybe your inadequacy as an Omega).
comment by AdeleneDawner · 2009-08-19T23:12:45.629Z · LW(p) · GW(p)
Interesting.
How did you come to those numbers? Do you really think Eliezer would be willing to post that for a bit less than $7? That seems a bit low, to me.