As discussed in this article in New Scientist,
researchers used Avita to create populations of identical “digital organisms” that initially were incapable of solving logical problems. But with each replication, there was a 20 per cent chance of a random mutation in “offspring.” This mutation altered the nature of the digital organism and in some cases resulted in one that could perform a logical operation. After 15,000 generations, the researchers found it was impossible for a population of digital organisms to solve the most difficult logic problems if that was all that the computer rewarded. But the outcome changed dramatically if the digital organisms lived in environments that would also reward them if they performed some simpler functions. In that case, the evolving programs were able to bridge the gap and eventually solve even the most complex logic problems.
“Our work allowed us to see how the most complex functions are built up from simpler and simpler functions,” says Lenski. Added his collegue Charles Ofria, “One of the beautiful aspects of this work is that it allows us to better understand how nature overcomes difficulties inherent in solving complex problems. We can then apply these concepts when trying to decide how best to solve computational problems we are faced with.”