Evolutionary theory in computer science has been unsuccessful
From EvoWiki
Contents |
Claim
Failure or shortcomings in experiments involving the application of evolutionary theory to the production or use of computer hardware of software demonstrates that evolution can not work elsewhere.
Source
- Rana, Fazale R., n.d., �Evolving� Robots Challenge Evolution in Facts for Faith, Issue 5 - Reasons to Believe.
Responses
- Computerized evolutionary models have proven effective. See [1] for the Avida software, which is also available for download. An analysis of the software and it's results indicates that computer models can be used to demonstrate evolutionary theory[2]
- The criticisms about evolutionary models are often directed at a Brandeis experiment in evolving concepts for machines via simulation of the evolutionary process. These are better viewed not as a simulation of evolution, but an application of evolutionary concepts to producing mechanical design. Though useful and interesting in that it did produce surprisingly sophisticated mechanisms, it was not a full-bore testing of evolutionary theory in the vein of the Avida software.
- There are programs that use computer-simulated evolution to solve certain tasks. One amusing and rather simple example of this is breveWalker, where an animal made out of blocks uses evolution to learn how to walk. (Suggestion: add more examples.)
- In terms of purely practical use in computer science, many of the fastest sorting algorithms in the world have come from genetic algorithms. See papers by Danny Hillis.
- Compare with: "Failure or shortcomings in experiments involving the application of evolutionary theory to the production or use of ICE CREAM demonstrates that evolution can not work elsewhere."
Fallacies contained in this claim
- Argument from Ignorance (evolutionary theory in CS is a developing field, it is too difficult to make conclusions of it's success or otherwise yet)
- False Analogy (evolutionary theory in CS is similar to Darwin's evolution, therefore the properties of the former must apply to the latter)
- Straw Man (if one can not attack evolution using scientific evidence, attack Darwinian evolution under the auspices of something else)
- Fallacy Fallacy (computer science evolutionists may be right in principle even if they make mistakes)
- Category Error (if it can't work one place, it can't work anyplace)
