LessWrong is currently deciding to do something that will have an estimated 90%+ chance of blowing up the home page. This decision is 90% as bad as just blowing up the home page.
What is the utility function that might cause LessWrong to take such a decision? I can’t come up with anything that fits other evidence. This has confused me in prior years but now there’s a prediction market I can quantify my confusion more easily.
If we calibrate so that “taking down the homepage” (redirects to 404 error or whatever it is) for 24 hours is of badness 1 (goodness = −1).
Then the “game over” page is of badness <1 (maybe 0.75?).
Then the whole red-button game is fun, and gives people (like both of us) something interesting to talk about, which is worth some goodness (-X badness). Plus the homepage might be down for less than 24 hours.
So they think : X >~ 0.75
IE The Lesswrong team appear to have decided that the fun of the game (playing/speculating) is worth more than a few hours of homepage plus the time spent implementing the game.
First observation: this represents a different set of values to those I infer from the 2020 post-mortem. Good, perhaps there have been updates.
Second observation: Games are more fun when the outcome is less certain, and currently it seems very likely to result in “game over”. The participant count is higher than 2020 and players are anonymous.
Modifying the rules to reduce the estimated chance of “game over” would thus increase the upside and decrease the downside.
There are also other games to play, keeping the fun but without the costs. How about a nice game of chess?
But the outcome IS uncertain. I want to know how low the karma threshold can go before the website gets nuked. There are other fun games, but this one is unique to LW and seems like an appropriate way of celebrating Petrov Day.
I wonder how it would change things if there was an additional rule: “the button will be taken offline after X hours, pulled from [publish the distribution], unknown to anyone but Ruby in advance”.
The 2020 and 2021 retrospectives did not emphasize fun as a terminal value for LessWrong admins, as I read them. If they now believe that blowing up the home page is fun then I understand the decision more. Looks like they will have that fun twice today.
Nit pick: There is a 90%+ chance it’s an exercise in collective inadequacy. Do you think that this proposal is positive EV because of the value of running such an exercise?
Given the probabilities involved, that implies that the net benefit of running an exercise in collective adequacy, is more than 10x larger than the cost of having the home page blown up. That seems large.
Also, simply increasing the karma cut-off would allow the same exercise to be run with a lower chance of blowing up the home page. Would this be so much less valuable than the current proposal, as an exercise in collective (in)adequacy?
Personally, on net, I think LessWrong shouldn’t create the button. I think it probabilistically destroys some real value and is symbolic of destroying real and even greater value. Plus, I think there’s symbolic value in not creating buttons that launch nukes.
LessWrong is currently deciding to do something that will have an estimated 90%+ chance of blowing up the home page. This decision is 90% as bad as just blowing up the home page.
What is the utility function that might cause LessWrong to take such a decision? I can’t come up with anything that fits other evidence. This has confused me in prior years but now there’s a prediction market I can quantify my confusion more easily.
If we calibrate so that “taking down the homepage” (redirects to 404 error or whatever it is) for 24 hours is of badness 1 (goodness = −1).
Then the “game over” page is of badness <1 (maybe 0.75?).
Then the whole red-button game is fun, and gives people (like both of us) something interesting to talk about, which is worth some goodness (-X badness). Plus the homepage might be down for less than 24 hours.
So they think : X >~ 0.75
IE The Lesswrong team appear to have decided that the fun of the game (playing/speculating) is worth more than a few hours of homepage plus the time spent implementing the game.
First observation: this represents a different set of values to those I infer from the 2020 post-mortem. Good, perhaps there have been updates.
Second observation: Games are more fun when the outcome is less certain, and currently it seems very likely to result in “game over”. The participant count is higher than 2020 and players are anonymous.
Modifying the rules to reduce the estimated chance of “game over” would thus increase the upside and decrease the downside.
There are also other games to play, keeping the fun but without the costs. How about a nice game of chess?
But the outcome IS uncertain. I want to know how low the karma threshold can go before the website gets nuked. There are other fun games, but this one is unique to LW and seems like an appropriate way of celebrating Petrov Day.
I wonder how it would change things if there was an additional rule: “the button will be taken offline after X hours, pulled from [publish the distribution], unknown to anyone but Ruby in advance”.
The 2020 and 2021 retrospectives did not emphasize fun as a terminal value for LessWrong admins, as I read them. If they now believe that blowing up the home page is fun then I understand the decision more. Looks like they will have that fun twice today.
It’s an exercise in collective adequacy.
Nit pick: There is a 90%+ chance it’s an exercise in collective inadequacy. Do you think that this proposal is positive EV because of the value of running such an exercise?
Given the probabilities involved, that implies that the net benefit of running an exercise in collective adequacy, is more than 10x larger than the cost of having the home page blown up. That seems large.
Also, simply increasing the karma cut-off would allow the same exercise to be run with a lower chance of blowing up the home page. Would this be so much less valuable than the current proposal, as an exercise in collective (in)adequacy?
Personally, on net, I think LessWrong shouldn’t create the button. I think it probabilistically destroys some real value and is symbolic of destroying real and even greater value. Plus, I think there’s symbolic value in not creating buttons that launch nukes.
Oh I support increasing the karma cutoff.
I do think that running such an exercise is valuable, if only because it allows us to learn things about our community.