Dembski Wakes Up, Smells the Steiners, Pushes Snooze Button

By Dave Thomas

I talked to Bill Dembski in person about my work on using Genetic Algorithms to solve Steiner’s problem way back in 2001. He didn’t “get” it then, and he still doesn’t!

Reacting to this news story, “Supercolony trails follow mathematical Steiner tree”, Dembski writes today that

Some years back, ID critic Dave Thomas used to tout the power of genetic algorithms for their ability of solve the Steiner Problem, which basically tries to minimize distance of paths that connect nodes on a two-dimensional surface (last I looked, he’s still making this line of criticism - see here). In fact, none of his criticisms hit the mark – the information problem that he claims to resolve in evolutionary terms merely pushes the design problem deeper … In ID terms, there’s no problem – ants were designed with various capacities, and this either happens to be one of them or is one acquired through other programmed/designed capacities. On Darwinian evolutionary grounds, however, one would have to say something like the following: ants are the result of a Darwinian evolutionary process that programmed the ants with, presumably, a genetic algorithm that enables them, when put in separate colonies, to trace out paths that resolve the Steiner Problem. In other words, evolution, by some weird self-similarity, embedded an evolutionary program into the neurophysiology of the ants that enables them to solve the Steiner problem (which, presumably, gives these ants a selective advantage).

Kudos to Dr. Dembski for this classic Goal-Post movement! The purpose of my original article was simply to move the discussion of Genetic algorithms beyond the ID “Dawkins Defense,” namely that all genetic algorithms suffer the “Weasel” flaw of needing the solutions to be incorporated directly into the fitness function.

Dembski’s response is remarkable in that it totally avoids the issues I raised. Just because ants can find ways for colonies to make efficient paths has no bearing on whether genetic problems can be applied without having solutions in hand already.

My original article on Steiner (Target? TARGET? We don’t need no stinkin’ Target!) showed that there are also physical methods for solving Steiner’s problem, including minimal-surface soap films.

If soap films can solve Steiner problems, why not ants? And this bolsters the Weasel defense, how?

My Skeptical Inquirer article from last year, “War of the Weasels: An Evolutionary Algorithm Beats Intelligent Design” has a nice summary of these Weasel Wars, including the marvelous story of UD’s software engineer, Sal Cordova, getting whupped by a Genetic Algorithm on an open-book design problem. The article posting is courtesy of Southern Methodist University’s Critical Thinking/Physics Class!

More: Panda’s Thumb’s “EvoMath” category.