If you add to your analogy that you have no knowledge of the existing codebase, and you have no idea what you're adding or modifying, then you understand my difficulty grasping the concept. Thanks for the reply though!
Essentially you start with a random program and iterate on it using an evolutionary model. The point of these algorithms is precisely that an understanding and knowledge of the code base is not required. All that's necessary is some way of measuring the program's fitness for a task.
Right, but if you had _billions_ of years, you could still write an amazing program, even if you have no knowledge of the codebase and no idea what you're modifying.
Well, my point was that it is not just time, but also the massive parallelism. Both combined is just such a phenominally mind-bogglingly large amount of exploration.
