Just as a thought experiment, let’s consider how Natural Selection might “learn” in a similar way to how our brains learn.

How do humans learn? Humans can learn by trial and error, from experience, from habit, from conditioning, or from playing or tinkering. The common thread for all learning is behavior coupled with feedback. By behaving in a certain way and gathering feedback about that behavior, we learn. Positive feedback reinforces and refines behavior, while negative feedback lets us know when behavior is wrong. Any system with behavior, feedback and the ability to apply that feedback to behavior, will learn. (Notice the cyclical aspect of this kind of system.)

Natural Selection initially determines behavior through random mutation. New genes come about by slight copying mistakes (essentially typos) in the genetic code. Since the mutations are random, the behavior will be random — at first. Then, positive feedback is given for behavior that helps, and negative feedback for behavior that hurts. Natural Selection actually collapses feedback, and the application of that feedback to behavior, into a binary step: either the gene survives and passes on the accumulated feedback and behavior, or it dies and collected feedback and behavior are lost. In the brain, when negative feedback is received, it is used to change the brain, where in Natural Selection negative feedback is death. Nature makes up for this inefficiency by distributing this win/lose game over a massive network of animals. Only a very select few survive, but they have a 100% positive feedback rating. Every single ancestor of theirs was a survivor. The large network of animals acts like a distributed brain and has the appearance of learning.