Indeed. Given a lack of transcripts being released, I give a reasonable amount of probability that there is a trick of some sort involved (there have been some proposals of what that might be, e.g. “this will get AI research to get more donations”), although I don’t think that would necessarily defeat the purpose of the trick: after all, the AI got out of the box either way!
As I understand it, that would violate the rules, and it would be appealing to the utility of the person playing the Gatekeeper, rather than the Gatekeeper. If there were actually an AI trying to get out, telling the Gatekeeper “You’re actually just pretending to be a Gatekeeper in an experiment to see whether an AI can get out of a box, and if the result of the test shows that the AI can get out, that will increase research funding” would probably not be effective.
Well, put it this way, if Eliezer had performed a trick which skirted the rules, he could hardly weigh in on this conversation and put us right without revealing that he had done so. Again, not saying he did, and my suggestion upthread was one of many that have been posted.
How do we know EY isn’t doing the same?
Indeed. Given a lack of transcripts being released, I give a reasonable amount of probability that there is a trick of some sort involved (there have been some proposals of what that might be, e.g. “this will get AI research to get more donations”), although I don’t think that would necessarily defeat the purpose of the trick: after all, the AI got out of the box either way!
As I understand it, that would violate the rules, and it would be appealing to the utility of the person playing the Gatekeeper, rather than the Gatekeeper. If there were actually an AI trying to get out, telling the Gatekeeper “You’re actually just pretending to be a Gatekeeper in an experiment to see whether an AI can get out of a box, and if the result of the test shows that the AI can get out, that will increase research funding” would probably not be effective.
You’re quite possibly right, and without access to the transcripts it’s all just speculation.
I don’t think we need the transcripts to discuss whether a hypothetical strategy would be allowed.
Well, put it this way, if Eliezer had performed a trick which skirted the rules, he could hardly weigh in on this conversation and put us right without revealing that he had done so. Again, not saying he did, and my suggestion upthread was one of many that have been posted.