Joe G
Posts: 12011
Joined: July 2007
|
| Quote (Febble @ Feb. 16 2014,06:26) | | Quote (Joe G @ Feb. 15 2014,18:08) | Genetic algorithms are goal-oriented. They are designed for specific purposes. For a GA to design an antenna, for example, all of the information for that antenna has to be programmed in and the offspring are compared to that. They employ a targeted search and cumulative selection to achieve a pre-specified result, ie the specification of the antenna required.
Dawkins’ “weasel” is unable to design an antenna because it isn’t designed to. Only GAs specifically designed to design an antenna can do so. Dawkins' weasel wouldn't have found the target sentence if that wasn't front loaded into the program. The antenna program never would have designed the proper antenna if the specifications for that antenna wasn't front loaded into the program. Got that, dipshit? |
Jo, the only way in which a GA differs from a "natural" evolutionary scenario is that the designer of the GA usually has a problem or set of problems that she wants to solve.
In nature we have no particular reason to think that the world was set up so that some Designer could find a better way of flying, or swimming, or invading the guts of small children, but it might have been.
But it doesn't matter, because in both cases, from the point of view of the population of organisms (virtual or biological) the problem it has to "solve" is simply how to leave viable offspring in the environment in which it finds itself.
The fact that in a GA the environment was set up by a designer to solve her own problem as a byproduct, and in nature it may not have been, is irrelevant.
The designer of a GA may have a goal, but the evolving population's only "goal" is to survive in that environment. If someone accidentally alters the fitness parameters of the GA, the population will evolve to survive in the new environment, possibly by evolving a maximally bad antenna, or one that only receives signals in some random set of bands. In fact one problem that GA designers encounter is that populations find ways of surviving and breeding in the artificial environments that do not solve the problem the designer wanted solving.
Just as setting up a natural environment with high trees in the hope of evolving a solution to the problem of flight might get you fruit-bats or flying squirrels, but might instead only get you organisms that are extremely good at grasping branches (e.g. spider monkeys), or bouncing unharmed when they hit the ground (e.g. hedgehogs).
For either a GA or natural evolution to result in a population that evolves functions that help its members survive in the current environment, all you need is an environment that has resources that can be accessed by enhanced functions, or threats that can be avoided by enhanced functions. It doesn't matter to the process or the population what those resources or threats are.
If you are a designer, you might want to set them in such a way that the evolving population will have to solve your problem (e.g. making a better antenna) in order to solve theirs (surviving and breeding).
Just as a farmer sets up his selective breeding program in order that the evolving population of cows solves his problem (improving milk yields) while also solving theirs (making the cut into the breeding pool).
In other words, the "goal" of a GA is not the "goal" of the virtual critters within the GA. You are confusing the two. The population of critters has no "goal" in either case, apart from leaving viable offspring.
That a designer has an ulterior goal in solving some problem of her own, and has set up the resources and threats of the environment so that by adapting to it, the evolving population is likely to solve it as a byproduct, is irrelevant to whether adaptive evolution occurs. It will occur whether or not it solves somebody else's problem or not. And as a GA designer, it can be frustrating when your own goals aren't met. It won't mean that adaptive evolution hasn't occurred - but it might mean that the adapting population found a solution to surviving in your carefully engineered environment that didn't happen to solve the problem you wanted solving.
BTW, Weasel is special case, because genotype, phenotype and fitness function are identical. It's just a toy. It's a perfectly good toy, and it illustrates the principle of adaptive evolution perfectly well (the evolving population of letter strings have to compete for "resources" in an environment in which those resources more available for strings that more closely resemble the phrase "Methinks it is like a weasel"), but it lacks key features of both natural evolution and practically useful GAs. |
Lizzie,
Just stuff it already. GAs are nothing like darwinian evolution. GAs have a goal and use targeted searches to achieve that goal.
Also neither AVIDA nor Tierra are GAs and neither of them model darwinian evolution.
Look Lizzie, Donald Johnson, PhD in computer science, wrote all about this and you are wrong, again, as usual.
--------------
"Facts are Stupid"- Timothy Horton aka Occam's Afterbirth
"Genetic mutations aren't mistakes"-ID and Timothy Horton
Whales do not have tails. Water turns to ice via a molecular code- Â Acartia bogart, TARD
YEC is more coherent than materialism and it's bastard child, evolutionism
|
