One reason for Eliezer not publishing the logs of the AIbox experiment is to avoid people seeing how he got out and responding, “ok, so all we have to do to keep the AI in its box is avoid succumbing to that trick.” This thread might just provide more fuel for that fallacy (as, I admit, I did in replying to Eliezer’s original comment).
I’m sure that for everything an AI might say, someone can think up a reason for not being swayed, but it does not follow that for someone confronted with an AI, there is nothing that would sway them.
It just occurred to me that Eliezer’s original stipulation that no chat logs would be released gives him an advantage. The responses of a Gatekeeper who knows that his inputs will be thoroughly scrutinized by the public will be different then one who has every reason to believe that his discussion will be entirely private.
Honest question: are you proposing we avoid discussing the problem entirely?
Personally, I think there is more to be gained here than just “how will an AI try to get out and how can we prevent it.” For me, it’s gotten me to actually think about the benefits and pitfalls of a transhuman AI (friendly or otherwise) rather than just knowing intellectually, “there are large potential benefits and pitfalls” which was my previous level of understanding.
Edit: I’ve modified the OP to include your concerns. They’re definitely valid, but I think this is still a good discussion for my reasons above.
Honest question: are you proposing we avoid discussing the problem entirely?
No, I just thought that it was worth adding that concern to the pot.
I take what I dare say some would consider a shockingly lackadaisical attitude to the problem of Unfriendly AI, viz. I see the problem, but it isn’t close at hand, because I don’t think anyone yet has a clue how to build an AGI. Outside of serious mathematical work on Friendliness, discussing it is no more than a recreation.
Discussing it makes people more aware of exactly how difficult a problem it is. That such discussions are entertaining merely permit the discussions to take place.
One reason for Eliezer not publishing the logs of the AIbox experiment is to avoid people seeing how he got out and responding, “ok, so all we have to do to keep the AI in its box is avoid succumbing to that trick.”
One reason for Eliezer not publishing the logs of the AIbox experiment is to avoid people seeing how he got out and responding, “ok, so all we have to do to keep the AI in its box is avoid succumbing to that trick.” This thread might just provide more fuel for that fallacy (as, I admit, I did in replying to Eliezer’s original comment).
I’m sure that for everything an AI might say, someone can think up a reason for not being swayed, but it does not follow that for someone confronted with an AI, there is nothing that would sway them.
I wouldn’t expect any effective real-life gatekeeper to be swayed by my ability to destroy one-sentence AIs.
It just occurred to me that Eliezer’s original stipulation that no chat logs would be released gives him an advantage. The responses of a Gatekeeper who knows that his inputs will be thoroughly scrutinized by the public will be different then one who has every reason to believe that his discussion will be entirely private.
Has someone else pointed this out before?
Honest question: are you proposing we avoid discussing the problem entirely?
Personally, I think there is more to be gained here than just “how will an AI try to get out and how can we prevent it.” For me, it’s gotten me to actually think about the benefits and pitfalls of a transhuman AI (friendly or otherwise) rather than just knowing intellectually, “there are large potential benefits and pitfalls” which was my previous level of understanding.
Edit: I’ve modified the OP to include your concerns. They’re definitely valid, but I think this is still a good discussion for my reasons above.
No, I just thought that it was worth adding that concern to the pot.
I take what I dare say some would consider a shockingly lackadaisical attitude to the problem of Unfriendly AI, viz. I see the problem, but it isn’t close at hand, because I don’t think anyone yet has a clue how to build an AGI. Outside of serious mathematical work on Friendliness, discussing it is no more than a recreation.
That’s pretty much my same attitude on the situation, as well. :)
Discussing it makes people more aware of exactly how difficult a problem it is. That such discussions are entertaining merely permit the discussions to take place.
He could post the logs of the games he lost.
Thereby giving a different reason for false confidence in boxing.
Can you elaborate, please?