It has been a century and a half since Darwin provided the first mechanistic explanation for the complexity of the living things we see around us. Only in the last 30 years or so have computational systems been employed to try out natural selection on complex artificial problems. There have been some successes, but the complexity of artificially evolved systems remains a very long way short of the complexity that is easy to find in biology. Why is this? Is our understanding of natural evolution missing something important? How can we improve our artificial problem solving methods to make them work better on large-scale complex problems?