We can harness the equivalent timeframes of evolution with simple GAs over quick runs (<1m), but we get nowhere near the results. What is the secret sauce?
I just tested one of my GA's and it's solving a logic problem with 3k pop, 30 children per gen, in 1-2k generations. If I make it 300 children per gen it solves in 150-300 generations. 1500 children is solving in 30-50 generatikons. Now imagine 5 x 10^30 pop with 10^30 children per generation, over millions of generations.