Not a neural network yet, but this IS a genetic algorithm. Baby steps.

1. Creation: It creates 20 random genes (strings) that are the same length as the target string (the one you input).
2. Fitness: It deteremines how "far" each is from the target string.
3. Mating: The two "closest" strings are mated (the first half of each is spliced with the second half of the other).
4. Mutation: Every generation, each of the 20 strings has a 50/50 chance of getting mutated (having one character changed up or down by one place. For example, an E might become a D or an F).
5. Evolution: Repeat from step 2 until the target is found.

To appreciate the randomness, try the same string twice. It will almost certainly take a different number of generations to find them.