Well, here is another citation.
The role of mutational and other biases in variation
In reality, mutation is not random in any sense other than in respect of being logically prior to selection (which is not a reason to call it "random"), and the idea of a bounteous Dobzhanskian "gene pool"-- the magic hat from which selection can pull any trick-- died years ago.
Although recent commentaries give the impression that the differences between classic neo-Darwinism and "evo-devo" are being resolved, we do not believe that this is the case. One cannot integrate these two contradictory views. Either propensities of variation exert an important shaping influence on evolution, or they do not. By implying the former, the "evo-devo" heterodoxy strikes at the root of the New Synthesis: its case for the supremacy of selection and its case against any possible internal causes of directionality. Far from being resolved, the controversy has not even been clearly recognized yet.
For the last two months, I have been working on a rather difficult problem to register high resolution images from multiple cameras together. Using the standard methods, it was taking around 5 minutes to process and that was way too long.
Since I love to read these E vs C threads, one day it dawned upon me. Could I evolve a transformation matrix which would register the multiple cameras?
Today I gave the new software it's first test. I was simply stunned and amazed at how well it worked. Not only was it accurate, but it reduced the processing time from the previous 5 minutes down to 30 seconds.
Not bad for a bunch of random mutations with a little natural selection mixed in. Perhaps I will try to implement the new information you provided tomorrow.