The following is my own peice of information to add to this debate.
It isn’t exactly proof of the theory of evolution, just evidence of natural selection acting on random variation as an algorithm for seeking solutions. Still, once it is shown that natural selection does pretty much what Darwin thought it would do, evolution doesn’t seem that implausible does it?
Anyway, my evidence is as follows: I wanted to solve a specific engineering problem - how to get the ideal trajectory that a rocket had to take to reach orbit. The trajectory is based on a complicated function designed to minimize the drag and gravity losses of the path. If it’s too shallow, the drag losses would make the path very inefficient. If it’s too tall, gravity losses would make it difficult to attain orbit. Rockets, being very fuel inefficient vehicles, don’t have much of a fudge factor for fuel inefficient paths - so if you want to get the most out of your rocket, finding the best path is important.
So what does this have to do with natural selection? Natural selection, is, in effect, how the program that solved my problem works. I parameterized the path angle at each point in time (a steering function). This is the “genome”. Then I formed a population of trial genomes, all initialized to zero degrees at all times (which is a straight vertical trajectory). Obviously it wouldn’t reach orbit. Then I set up a loop where each genome would be tested by running a crude ascent simulation for the rocket (drag, gravity, ect). Afterwards, the performance of the rocket was ranked according to how close to a circular orbit tangential velocity it got, as well as which path got to a higher orbit. There were some things that went into it.
What was important was that I never specified anything about the solution other than what I wanted it to do (get to the highest possible circular orbit using a given rocket). I didn’t have to know how it would do it.
Anyway, then I would select the best genome out of the population, repopulate the population space with copies of that genome. I would then apply random mutations to the genome parameter - adding or subtracting a random value from a random position. Then I would re-iterate the simulation.
Each generation had a best performing genome. Even though the genomes were mutated randomly, due to the fact that the population was large (about 100), the fitness scores of that genome almost exponentially approached a maximum value, and the parameterized flight path approached a solution that solved the problem.
So, it worked. Natural selection can produce a solution to a design problem. What do you think?