[eu-gene] need advice on a framework design
list at philipgalanter.com
Sun Jan 29 19:39:54 GMT 2012
On Jan 28, 2012, at 4:15 PM, David Hart wrote:
> > If you had an evolutionary system to design airplanes, but the airplanes fell from the sky, I don't think you would refer to that evolutionary system as "working."
> On the contrary, as long as the fitness function measures flight time, and over time the flight times get longer, it *is* working, it is not really a matter of opinion or debate. Just because you want it to fly, doesn't mean that it will, and it also doesn't mean evolution is a failure. If we tried to design airplanes via artificial evolution, we would without a doubt run through many crashing designs before arriving at one that flies. Turns out, direct design actually did produce a bunch of crashing designs before we got to one that flies. We went through fewer crashes than we would have if we'd used evolution to make aircraft.
> Natural evolution does exactly this. How many generations would it take of us jumping off cliffs before we evolved flight? We'd go extinct after 1. Does that mean natural evolution is a dead end? We humans can't fly just because we want to. We haven't evolved invincibility. We can't breathe water or fire. Our skeletal systems are weak and prone to failure. We have adapted only to the existing natural forces of our species, and there is a long list of failure up to this point. If you put humans in any other context than our current environment, we can't survive.
David, you seem to be intolerant of language having words that are overloaded with multiple meanings. What I'm pointing out is that *as a programmer* you might say that an evolutionary system is working because it's technically complete and functioning as expected. But *as an artist* I would say it isn't working until it helps me make art of value. You can respond "keep going, eventually it will generate art of value." That's fine. But as an artist I'll say "OK, it isn't working yet, but it may in the future." This is not an illegitimate use of the language. It is legitimate to allow individuals to have different goals and thus different interpretations as to what "working" means.
> If you want to play with or opine on artificial evolution, it would be a good idea to better understand this fundamental point.
You assume too much.
> > Similarly an artist's evolutionary system has to generate good art or else it's not working either...at least not in an artistic sense.
> Must a paintbrush generate good art? Must photoshop generate good art? If they don't generate good art does that mean they're not working? Why must artificial evolution generate good art for an artist?
The difference is that an evolutionary system claims a degree of autonomy and the tools you mention do not. In the sense of "generative art" a paintbrush doesn't generate anything at all.
> > Perhaps a computer scientist could find love for an evolutionary system that designs fatal aircraft or bad art. Not me.
> You are using criteria that are unreasonable and misunderstand what artificial evolution is supposed to do, in order to claim its lack of utility. You are expecting evolution to give you the correct answer, and yet you aren't holding any other medium or technique to the same unrealistic standard. Humans have created lots of fatal aircraft and lots of bad art, through a wide variety of means. If you can't tolerate any risk of failure, you have exactly 0 options.
David, I'm sorry, but I understand exactly what artificial evolution is and how it operates.
But what it's *supposed to do* depends on who's doing the assessment. If you are a research programmer, for example, for an investment firm and you have a genetic representation, and a fitness function, and a method of creating variations, and the system is just clicking along as a programmer you might feel like it's working. But if it doesn't generate a profit the investors are going to say "hey, this evolution stuff doesn't work, stop using it to invest my money!" And they would be right to say so although you could plead for more time. And at best they would say "OK, you have X amount of time to make it work." And again, there would be nothing wrong with their use of the language and the word "work."
And note that I've never said that evolutionary systems can't generate art worth making. There are already, in my mind, examples that prove it is possible because it has already happened. So to say I'm unfairly indicting evolutionary computing in art is simply not true, because I'm not indicting evolutionary computing at all.
All I've said is that from what I've seen the typical Sims-like evolving expressions system (1) has an aesthetically limited phenotype space and (2) I don't think useful art is left to be found in that phenotype space.
This is a very specific observation and not at all an indictment of evolutionary computing as applied to aesthetic ends.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the eu-gene