Can evolution generate complex, specified information?
Read that.
I confess that I did not read the whole paper. It's written in a language that's far too technical for me. Are there any freely available summaries of this paper? Was this paper reviewed in any math journals?
From what I can understand NFL seems to be saying that for a certain kind of search problem there is no search method that out-performs any other search method. Is that right?
Helen, You relayed the idea that "artificial evolving systems such as genetic algorithms can generate novel, unprecedented and sometimes quite elaborate solutions." The issue here centers on exactly what is meant by "novel", "unprecedented", "elaborate", etc.
These algorithms are essentially search strategies implemented in software. The follow the general "evolutionary" pattern of trying many variations, evaluating the results, discarding unpromising results (the counterpart to natural selection), and repeating with new starting points. Chess programs and other game playing programs can similarly employ this kind of search to look for promising lines of play. This can be used effectively in certain kinds of situations (e.g. a new shape for an antennae), but there are qualifications.
Are these solutions "novel" or "unprecedented"? Well, they certainly can be different and it may be that no human had yet proposed to use that particular design. But those solutions are bound within the range of possibilities anticipated by the programmers.
To start with, the nature of the variations are specified by the programmers. They decide in advance, for example, whether to vary the number of arms on the antennae and/or the angles and/or the arm lengths, and so on. Also, the software cannot evaluate any possible solution except through the evaluation function provided by the programmers. So the evaluation function must anticipate and be able to cope with the range of possible solutions the variations could create. It cannot go beyond this or it could not evaluate a proposed solution.
Dembski's position does not exclude the possibility of an information processing system converting or processing information, as these programs are doing. They are built and designed by programmers specifically to behave in these ways. The specified complexity that is output is the expected and intended result of the specified complexity that went into the designed search strategy. Consider "design" in the following book title and it's description.
"The Design of Innovation; Lessons from and for Competent Genetic Algorithms,"
by David E. Goldberg, University of Illinois at Urbana-Champaign, USA
"The Design of Innovation illustrates how to design and implement competent genetic algorithms -- genetic algorithms that solve hard problems quickly, reliably, and accurately -- and how the invention of competent genetic algorithms amounts to the creation of an effective computational theory of human innovation. For the specialist in genetic algorithms and evolutionary computation, this book combines over two decades of hard-won research results in a single volume to provide a comprehensive step-by-step guide to designing genetic algorithms that scale well with problem size and difficulty."
I recently read a summary of a computer-science paper (which contained absolutely no references to biology, Darwinism or the origin of life) that artificial evolving systems such as genetic algorithms can generate novel, unprecedented and sometimes quite elaborate solutions.
I'd like to know what y'all think about this - my understanding was that complex specified information could never evolve. But can it? What does Dembski's take on information theory have to say about the claims of these computer-science types?
Thanks
---
Helena Petrovna Blavatsky
login or register to post comments | 0 points