Why Evolution sucks

Posted by
April 20th, 2010 8:36 pm

Have you ever worked with ANY program involving evolution? They all take hours at best, weeks at worst, just to run long enough to get interesting results, let alone to develop.

Way back I used to play with SodaConstructor, an applet where you make 2D moving creatures out of springs and muscles (which are springs that expand and contract according to a sine wave) joined at the ends. A related applet was SodaRace, where you built a track for these creatures to race on, and tried to build the one that got to the end the fastest. Someone made a program that took a creature and randomly varied various aspects of its construction, put a bunch of these mutants on a track, and raced them, picking the fastest ones and repeating this in a process of evolution. It took hours just to get something remotely interesting, and days to come up with a maybe kinda cool result.

Same story for a bunch of life simulators that I’ve tried that you can google for. Usually the story is I run it, see a bunch of weird creatures move around randomly being stupid, and maybe if I leave it running overnight the next day I’ll see a bunch of one creature that moves around randomly, in a way that happens to not consistently lead to death. It’s watching paint dry.

There are several problems with simulating evolution:

  1. It’s hard to come up with a digital environment that will support interesting evolution mechanics without simulating the trillions of atoms and chemical interactions that evolution has to work with. An arbitrary digital environment like a SodaConstructor applet will always have quirks that mean the most efficient creatures – the one most likely to be selected – also use the cheapest tricks and tactics, which includes bugs and glitches if there are any. Evolution will by definition test the very limits of its environment, so the tiniest glitch or imbalance in your simulation will completely throw off the quality of the results, if you ever get any.
  2. No one has yet devised an evolution algorithm that knows which mutations are likely to be effective, so random evolution goes at a snail’s pace. Of course, which mutations will be effective depends entirely on the environment; for complex environments there might never be an algorithm any more efficient than Real Life(tm). This compounds the problem of making a suitable environment in the first place, because it means it could take days between test cycles developing a system that’s already extraordinarily complex.
  3. The results are boring. It’s just the same kind of life (or computer programs) that we already see all around us, except inefficient, inelegant, and ugly, because of their primitive and largely random nature.

Evolution as a theme isn’t necessarily bad. There are other perfectly valid, doable interpretations of it. I’m just saying, if you voted for Evolution thinking you were gonna make an evolution sim, please reconsider.

12 Responses to “Why Evolution sucks”

  1. jplur says:

    Nice rant, but like you said there are a ton of interpretations on that theme.

    I tend to down vote themes that use a particular programming trick, i.e. boids or fluid dynamics.

    I also down vote cute themes, like kittens using baked goods as weapons, because I don’t want to barf on my keyboard.

  2. Deepflame says:

    I downvote evolution on grounds that it is way too much work for a 48-hour timeframe. 😛

  3. billknye says:

    The only interesting part of evolution would be, 48 hours to develop the program, maybe let it run for a couple weeks and compare results then.

  4. Codexus says:

    Not to mention that a successful entry could end up creating SkyNet and kill us all. Evolution is bad, mmmm’kay?

  5. bluescrn says:

    I’m another one that’s been downvoting Evolution since LD48 had voting…

    (Agree with Deepflame, to do the theme justice would take well over 48hrs. Could be an interesting theme for a different/longer compo, though)

  6. Tenoch says:

    There are tons of doable games involving evolution, in 48h. I agree about your view on using evolution simulation (like genetic algorithms) inside a game, it would probably suck. But Pokémon has evolution in it. Warcraft 2 has evolution in it (units and buildings). Evolutionists against creationists FPS. Primordial soup RTS. Whatever.

    We know it’s never gonna pass as a theme. But it’s the hope that makes us live!
    Evolution upvoters, unite!

  7. snowyowl says:

    Agree with Tenoch. Nobody said you have to use genetic algorithms. Spore has evolution in the Creature Stage, and it’s an awesome game.

    But yeah, genetic algorithms are overrated.

  8. Jonny D says:

    Never played EVO (for SNES), huh? You’re stuck in the weird idea of evolution that scientists have. We’re game developers. We can do better than that! Evolve literally means “to fly out of”. Take that however you want. Things really just need to change (however the player might control it) over time (however short or long) to satisfy the theme, and many, many interesting games can come of it, even in 48 hours. Evolution + 1!

  9. Jpfed says:

    Chemists say that a sample “evolves” gas when it releases gas. Just in case someone was interested in taking one of the most popular iPhone games out there and making it cross-platform.

  10. rob says:

    Evolution is a great theme. Control Charles Darwin as he explores uncharted islands and collects data to verify his theory.

  11. SonnyBone says:

    Yeah, see?

    Your game doesn’t need to EVOLVE in order for the game to be about evolution.
    In fact, nothing needs to evolve.

    Just make a tactical strategy game where you have to take over certain areas of the USA by adding Creationism to the school curriculum.

    (you play the bad guy, that’s the twist)


Leave a Reply

You must be logged in to post a comment.

[cache: storing page]