Has anyone ever suggested that the algorithm for intelligence might just be a tractable solution to an NP-complete problem, and that we live in a P=NP (or approximate) universe? I don’t believe this, because it doesn’t map to anything I understand about AI or the human brain, but it’s an interesting idea and I’m sure someone has pursued it before.
That would be shocking in a very fun sort of way…after breaking all cryptography on the internet, what sort of practical applications would such a discovery have?
Well, in this universe, NP and AI are secretly the same problem, so we’d get all the benefits of AI (singularity?) plus all the benefits of solving NP-class problems at the same time. I can’t think of an application of NP-class problem solving that would catch notice compared to AI, though. We already solved protein folding. Reduced carbon emissions from men who happen to be in sales and do a lot of travel, perhaps.
Protein folding is nothing compared with a general solution to finding proofs up to any given length if they exist or confirming that they don’t, or optimal solutions to virtually every scientific problem that can be formulated in any suitably verifiable way.
If solving NP-complete problems in polynomial time is the hallmark of intelligence, then humans are as dumb as rocks and any actual AI will blow us out of the water.
Has anyone ever suggested that the algorithm for intelligence might just be a tractable solution to an NP-complete problem, and that we live in a P=NP (or approximate) universe? I don’t believe this, because it doesn’t map to anything I understand about AI or the human brain, but it’s an interesting idea and I’m sure someone has pursued it before.
That would be shocking in a very fun sort of way…after breaking all cryptography on the internet, what sort of practical applications would such a discovery have?
Well, in this universe, NP and AI are secretly the same problem, so we’d get all the benefits of AI (singularity?) plus all the benefits of solving NP-class problems at the same time. I can’t think of an application of NP-class problem solving that would catch notice compared to AI, though. We already solved protein folding. Reduced carbon emissions from men who happen to be in sales and do a lot of travel, perhaps.
Protein folding is nothing compared with a general solution to finding proofs up to any given length if they exist or confirming that they don’t, or optimal solutions to virtually every scientific problem that can be formulated in any suitably verifiable way.
If solving NP-complete problems in polynomial time is the hallmark of intelligence, then humans are as dumb as rocks and any actual AI will blow us out of the water.