N.Wells
Posts: 1836 Joined: Oct. 2005
|
Quote (GaryGaulin @ June 18 2015,00:05) | And just when you would think it could not get any worse for the Darwinian paradigm Larry Moran just made it known that Evolutionary Algorithms do not really model "evolution" either:
Evolutionary algorithms are really adaptation algorithms http://sandwalk.blogspot.ca/2015.......ly.html
Science teachers are now being called into action to help defeat a misconception that leaves the Darwinian "evolution" paradigm with no reliable "model" at all. The reason for that is from not having an easily programmed model providing a framework that neatly sorts out all (intelligent or not) behavior systems found in biology to physics. |
And you dig your hole deeper again!
Larry said, Quote | I think we all know that "evolutionary" algorithms are based on natural selection and we all know that there's more to evolution than just adaptation. .... In one sense, it's okay to conflate "evolution" and "adaptation" in computer science but if that error reflects and perpetuates a fundamental misunderstanding of the nature of real biological evolution then perhaps it's time to rename these algorithms "adaptation algorithms." |
"Evolutionary algorithms" are in the field of computer science, not evolution. We and Larry have all been clear on this. (Avida and similar evolutionary simulations are a separate matter: Larry is not discussing them.)
I had slightly earlier said, Quote | "Evolutionary algorithms" are inspired by evolutionary processes and use reproduction, mutation, recombination, and selection to optimize solutions progressively and reiteratively. They are not biological evolution nor are they attempting to model biological evolution*. They are simply inspired by it. |
This is 100% consistent with what Larry is saying (Larry didn't spell out reproduction, mutation, and recombination because they are implicitly included in natural selection and adaptation). He contradicts what you have been saying. Do you see the words "genetic drift" and "speciation" in what I wrote? How about "not biological evolution"? Why do you think I specifically excluded genetic drift and speciation, and said "not biological evolution"? Because genetic drift and speciation are not necessarily part of evolutionary algorithms*, which is part of what makes "evolutionary algorithms" not biological evolution, which is identical to Larry's point. They are not trying to model biological evolution* nor any parts of it, nor are they trying to test it or prove it. They are merely inspired by it and use some aspects of it.
* I think it is correct to say that Avida and a few similar simulations use evolutionary algorithms, and they are indeed attempts to simulate evolution or parts of evolution properly. I'm not sure whether all of them use genetic drift, or are modelling other subsets of evolutionary processes. It's fine to model a subset of something (you could for example create a model of mutational processes that does not go on to cause evolution or modify any phenotypes, as long as it is not presented as a model of the whole system. As a shining example of that sort of bad practice, you have a model of foraging in poorly modeled sort of an insect and are presenting it as model of the emergence of intelligence.
Quote | From experience I know that without it you're stuck trying to combine a large number of existing models together then soon need a supercomputer to run them all at the same time. The NAS pdf I quoted is describing the same thing. |
Your experience is wrong-headed and irrelevant because you are confusing conceptual models (think of then as "schemes for understanding"), which is what NAS is talking about in that paragraph, with physical and computer models (think of them as reductionist demonstrations of the workings of the actual processes that make up the system, in order to demonstrate how typical results come about).
Quote | It's like having a couple dozen pieces of a 1000 piece puzzle showing the Intelligent Causation illustration that covers all else with how one behavior level has to connect to the next and the fundamental circuit of each.
The model and theory I am defending has the luxury of needing to remain as simple as what is shown in the illustration. That way the theoretical framework accepts and sorts out all new discoveries, without being changed by them. It only needs to stay the same, to carry on. | That's just delusional wrong-headedness and word salad from the outset.
Edited to add: From Judmarc at Sandwalk, Quote | A friend of mine developed "evolutionary" algorithms for the US Department of Defense "Star Wars" program. The degree of selection versus "neutral mutations" permitted can be adjusted. Therefore I wouldn't consider the quote in italics above to be completely accurate. | , so evidently some evolutionary algorithms do include drift.
|