Artificial intelligence has nothing to do with statistics

CommentaryZDNet reports that Jeff Hawkins, who created the Palm handheld, is forming a company that will specialize in systems able to recognize patterns, make predictions about familiar phenomena and in general function like the human brain.

Jeff Hawkins’ approach is insightful, I think, except on one strike: There’s not a chance in hell that the generally accepted model of infant language acquisition makes any sense. I’ll write a lengthy critic on why this statistic-driven model is meaningless in a more lengthy column some other time. For future reference however:

  • Bayesian statistics — like any other mathematical model — assumes variable independence; whereas, cognitively speaking, there is no such thing as independent variables
  • It is common for people like you and I to statistically misjudge situations, i.e. you and I forget that, statistic-wise, p(AB)<=p(A)p(B); I see no reason for an infant to be any different

Artificial intelligence will just be for another time.