‘How do you get to Carnegie Hall?’ ‘Practice, practice.’ Here’s a different way: start anywhere you like and take a step at random. If it’s a step in the right direction, I’ll say ‘warmer’; in which case repeat the process from your new position. If I say ‘colder’, go back a step and repeat from there. This is a kind of procedure that they call ‘hill climbing’ in the computer-learning trade (hence, I suppose, the title of Richard Dawkins’s new book). It’s guaranteed to get you where you’re going so long as the distance between is finite. (And so long as there are no insurmountable obstacles or ‘local maxima’ in the way: nothing is perfect.)

LRB 18 April 1996 | PDF Download