Joined: Sep. 2006
|GilDodgen: The “evolutionary strategies” to which you refer are simple trial-and-error algorithms that have a well-defined goal, and carefully crafted code that supplies well-defined heuristics which ensure that successive approximations to the goal can be reached with the computational resources in a reasonable amount of time.|
This is the exact antithesis of the Darwinian mechanism in biology at every step:
1) The algorithm is designed with foreknowledge of a goal.
Not all evolutionary algorithms have a singular goal. Evolutionary algorithms are quite adept at traversing complex and multivariate landscapes, with diverging populations exploring different areas of the landscape.
|GilDodgen: 2) The code is designed and optimized by computer programmers.|
Of course a computer model is designed by programmers. That's the whole point! Scientists carefully collect data and determine their interactions in order to simulate them with the computer.
GilDodgen is still having troubles with understanding the distinction between a model or abstract interpretation, and the thing being modeled. Considering how he was embarrassed by this previously, it would behoove him to try and grasp the concept.
|GilDodgen: 3) The hardware on which the code runs is designed.|
A computer modeling weather doesn't have to be out in the snow to accurately simulate a winter storm.
|GilDodgen: 4) The intermediate goals are predefined and contrived to be within the reach of the search strategy at every step. The intermediate goals are also given scores to evaluate the closeness to the goal numerically.|
If the intermediate scores were related to some future goal and not to current fitness, then this is a valid objection. But evolutionary algorithms do not require this sort of foresighted intervention.
Tard Acquisition and Repository Department