Pete, generally, evolutionary algorithms are algorithms that work with some "blind" (usually small) changes and test whether the results are somewhat better than before. If they are, start from there. In the end, you probably do not have an absolutely optimal product, but often much better than you could get by any other means.

The product can be anything, including - and that is the point - another computer system, which can then be regarded as "indirectly designed, directly evolved". The design part consists of the fact that the performance in each step is tested according to the designer's explicit criteria - whereas in "real" evolution, mere survival suffices for success.

Given the fact that a good eye is a gigantic advantage, the evolutionary pressure on it is accordingly high. Dawkins's point, presumably not quite serious, was that an almighty designer could have done even better. The general problem is that nobody could know the designer's criteria; nature itself never allows for verdicts like "good" or "better", beyond the fact of survival (which is actually a poor criterion for quality, given that many marvelous creatures became extinct because of changes in the environment).

In fact many people have the feeling that God designed the world "indirectly" in the above sense, including evolution, so that it appears to us as if by blind chance.

We must keep in mind that evolution theory of living organisms is based on genetics, much more complicated than simply effecting blind small changes (mutations).