Text Compression as a Test for Artificial Intelligence

摘要

It is shown that optimal text compression is a harder problem than artificial intelligence as defined by Turing's (1950) imitation game; thus compression ratio on a standard benchmark corpus could be used as an objective and quantitative alternative test for AI (Mahoney, 1999). Specifically, let L, M, and J be the probability distributions of responses chosen by a human, machine, and human judge respectively to the judge's questions in the imitation game. The goal of AI is M=L, the machine is indistinguishable from human. But the machine wins (the judge guesses that it is human) when H_J(M)