From "Nature, Design and Science", on Behe:
Michael Behe in Darwin's Black Box has been widely misunderstood ... as holding a [God of the] gap account for what he calls "irreducible complexity." His conclusion from that is not that nature could not produce irreducible complexity, but that if (perhaps since) nature has produced it then Darwinianism (as gradualist and selectionist) is simply wrong. (Behe does not even necessarily reject common ancestry - that is not really his issue.) I do not believe that Behe wishes to rule out supernatural intervention, but it is not essential to his more basic contention, which is that certain types and degrees of complexity (irreducible complexity) in and of themselves constitute evidence for design regardless of how they were produced.
(Notes on Design in Nature, p.187)
And on Dembski:
... imagine that we have a complete causal explanation back to the Big Bang of the present moon crater jumble, an objective measure of inherent complexity of the jumble array, and that we know the probability of precisely that jumble being produced. Now, without changing any of that, imagine that we discover that the present jumble in actual fact spells out "John 3:16" in Martian script. With no changes in causal history, degree, or type of inherent complexity, the special mind correlativity now visible gives the case a vastly different complexion. On the other hand, suppose that human written language had developed in totally different directions, so that the inscription "John 3:16" carried no meaning whatever. In that case, nature's generating of that pattern in moon craters would have generated no surprise at all (or at least much less). Obviously, the unexpected match between our inscripted language and nature's production is a key element of the situation. This is related to the concept of specification in William Dembski's The Design Inference. It seems to me, however, that Dembski never gets beyond a purely formal indication of specification, which leaves the term without any of the content necessary to support actual design conclusions.
(Notes on Design in Nature, p.187)
Which seems to be the objection that many other people have to Dembski - or one of them, to be more precise. In defence of Dembski at this point, I think it may be worth saying that he has regarded his work hitherto as representing a mathematical foundation for more applied investigation into systems - biological or otherwise - and it is possible to see at least in principle how the idea can be applied. Indeed, I started doing just that for both biological systems and random texts, but my lack of time and my limited grasp of the maths somewhat hampered me. Nonetheless, I would suggest that, given Dembski's rules about probability boundaries,
my proposition that a gene must initially code for a protein which is less than 7000 amino acids long is a non-trivial example of a prediction based on his theoretical work.
