[tortoise]

For better or worse, genetic programming and other evolutionary systems have limited application in making interesting things happen on standard computers. The problem is that you have to severely constrain the domain or your application will wonder off and do things that have no application to the problem you are actually trying to solve. To make matters worse, standard computational models don't do nearly as well as biological or non-biological chemical systems in being even remotely scalable enough for interesting evolution in a relatively unconstrained domain. The net result is that you can do interesting evolutionary work in chemical systems but not on computer systems, so lacking practical chemical computers genetic programming has been shelved for the most part. In other words, an engineering problem.

[js1138]

In other words, an engineering problem.

A couple of good patents in the restricted domain of electronic circuitry, and I suspect the engineering problems will be considered worthy of attention. There are zillions of design problems where the materials and desired outcomes are constrained.

The abstract I posted does not mention that the circuit designed by genetic programming is not fully understood, even though it works. I would assume that this meets a primary goal of artificial intelligence, the production of useful and novel designs that exceed the abilities of the program creator.