Peter Merholz recently linked to Pharyngula’s takedown of Ray Kurzweil:
Kurzweil knows nothing about how the brain works. It’s design is not encoded in the genome: what’s in the genome is a collection of molecular tools wrapped up in bits of conditional logic, the regulatory part of the genome, that makes cells responsive to interactions with a complex environment. The brain unfolds during development, by means of essential cell:cell interactions, of which we understand only a tiny fraction. The end result is a brain that is much, much more than simply the sum of the nucleotides that encode a few thousand proteins. He has to simulate all of development from his codebase in order to generate a brain simulator, and he isn’t even aware of the magnitude of that problem.
Kurzweil may well be wrong, but he’s not stupid (well, if he’s stupid he’s not as stupid as the above would make him seem): he’s not making an argument about simulating gene expression, but a totally separate argument based on Kolmogorov complexity, that is to say, what is the shortest algorithm you can use to reproduce the behavior. Simulating gene expression is a problem which everyone knows is incredibly hard, and may be impossible in general, as it requires the running of exponentially complex quantum field calculations — but that’s really not what Kurzweil has in mind here. He’s saying that an upper bound of the Kolmogorov complexity would have to be the genetic information required to generate the brain. Obviously he’s not saying that in 10 years we would literally build a gene expression simulator that could take the genome and generate a functioning brain. If classical processes can be used to simulate gene expression, then Kurzweil would certainly be right, at least in principle: the genetic information would be an upper bound of the minimum size of the algorithm needed to simulate the brain, regardless of which algorithm you use (and obviously Kurzweil imagines you’d probably use very different algorithms than nature uses).
He may be wrong, however, just because the generation of the brain may rely on quantum field effects which might allow for compression beyond that which a classical computer is capable of (i.e., it’s known that biological systems can take advantage of quantum effects, for example, a recent paper showed that plants take advantage of quantum computation). Quantum computers are of course capable of computational feats that would defy a classical computer, so his estimate could end up being wrong for that reason.
And even if he is correct, he may be grossly underestimating how long it might take for human beings to build algorithms with sufficient “compression” to generate brain-like behavior. But my main point is that most of Pharyngula’s blog post is beside the point, interesting as it is, because it attacks an argument Kurzweil is not making; i.e., the post is conflating algorithmic complexity with the difficulty of simulating gene expression, two totally different things.
permalink |