Nonsense Preferred to Nonsense
There are a few other points worth making about how different the Weasel game is from anything in the real world.
For one, the game lets you start from any string whatsoever, and it will move from that string to the final result in roughly the same number of steps every time. Say the starting string is YYYYYYYYYYYY... The program generates a batch of mutated copies and finds that the best fit among them is YEYYYYYYYY... But what makes this an improvement over the piece of junk it replaced? It's exactly as junky as before — no function, no meaning, nothing. It's kept only because the E happens to sit in the same position the E sits in "METHINKS." That's nonsense being preferred over nonsense, for a reason that has nothing to do with anything real. And yet the program treats it as the next foundation to build on, as if something meaningful had happened. It hasn't. Design a program to accomplish a specific task, and it will generally get the job done, unless the programmer is incompetent. Does that matter here? Enormously — because the same thing keeps happening, step after step. Each new piece of nonsense becomes the platform the next piece of nonsense is built on, all the way to the point where the program finally succeeds. That demonstrates the point Dawkins wanted to make. It is also about as far from biology as it's possible to get.
And this holds true almost no matter how you start. At any given step, one small change — one letter altered — has very little chance of conferring anything resembling a benefit in the ordinary sense of that word. So the program spends most of its time locking in one meaningless change after another, long before anything resembling a real benefit enters the picture, and it still finishes successfully every time.
It's worth being precise about what is meant here by "locking." The people who build and maintain these programs are right to point out that nothing is technically frozen in place. A letter that matches the target is not made immune to further mutation — it remains just as changeable as any other letter, every single generation. But that technical fact doesn't actually help. In any given early generation, there are far more still-wrong positions than correct ones, so the ordinary churn of mutation is much more likely to fill in one more blank than to undo a letter already sitting correctly. And there's a second factor doing even more of the work: the program doesn't generate just one mutated copy each generation — it generates many at once, and keeps whichever one scores best. If even one of those copies holds onto everything already correct and also picks up something new, that's the one that gets kept, and all its competitors are discarded. A correct letter only has to survive in a single copy out of the whole batch to be preserved. Put those two things together, and the result is a lock in every practical sense, even though no rule anywhere says "protect this letter." Our own simulation, later in this chapter, simply builds the lock in openly.
Nothing resembling this happens in real life. There is no mechanism, anywhere in biology, that prefers one meaningless string of junk over another meaningless string of junk. If anything resembling a lock exists in the real world, it exists only where a change confers some real, measurable advantage — something that actually helps an organism outcompete its rivals for survival or reproduction. And there's one more difference sitting underneath all of this: in the program, everything that isn't the chosen copy simply disappears. The winner is the only thing left standing, with no rivals, no competition, nothing else to contend with — and it's immediately available, at full strength, to produce the entire next generation's batch of candidates. Real organisms don't work that way. A beneficial mutation doesn't erase every competitor the instant it appears; it has to survive alongside all of them, spread gradually, and slowly outcompete a population that, for a long time, still vastly outnumbers it.
Last revised 2026-11-19