The trouble with AI

This article (spotted at Brains) is an excellent reminder of the problems with Artificial Intelligence, a welcome dose of reality against the naive triumphalism of people like Ray Kurzweil.

When the author discusses the connectionist approach, he touches on the intractable complexity that would plague the development process of a connectionist brain:

If each synapse were handled by the equivalent of only a single line of code, the program to simulate the cerebral cortex would be roughly 25 million times larger than what’s probably the largest software product ever written, Microsoft Windows, said to be about 40 million lines of code.31 As a software project grows in size, the probability of failure increases.32 The probability of successfully completing a project 25 million times more complex than Windows is effectively zero.

He doesn't mention potential ways to sidestep that problem. For example: evolving a population of complex artificial brains from a population of much simpler ones, through the use of random mutations and artificial selection. Conceivably, AI could be achieved by that method without the need for humans to understand in every detail how the evolved brains work, or how to design them explicitly.

Of course, that approach would face severe problems of their own. As the article mentions, simulating a single human brain would require a humongous amount of computing power. Simulating vast populations of them through innumerable generations would require much, much more. And then there are the problems of finding the proper encoding of the "genomes", the proper mapping between "genotypes" and "phenotypes", how to measure fitness, and so on.

Interestingly, the author does not claim that AI is impossible in principle. He never mentions arguments like Searle's The Chinese Room, or the ones spoused by Penrose involving non-computability. (Myself, I think the virtual mind/systems reply is a very convincing answer to Searle. I don't fully understand Penrose's argument, but I suspect the non-computable, Gödel's Theorem-defying aspects of the human mind he talks about are not really there.)

Edit: See also this post on Gene Expression.