I take it you haven't read Dembski's paper on specified complexity? If you can find the holes, I'd sure appreciate it if you could show them. Here's the URL: intellegent design as a theory of information.
I've read some of his other papers, which share much with this one. To address the problems with his paper point by point would require a paper nearly as long.
The short version is that he "proves" mathematics primarily with analogy and anecdote rather than rigorous mathematics. He misuses information theory and must only have a passing familiarity with it. The most annoying thing is that he frequently uses the "common" definition rather than the mathematical definitions of the terms used in information theory to prove his point. As you may know, many of the core terms used in information theory have mathematical meanings that in practice are quite different than the common definitions understood by most people.
Dembski basically seems to talk a very smooth pseudo-math, and capitalizes on the fact that people don't understand the difference (or that a difference even exists) between the mathematical definitions and the common definitions of the words he uses. Since he is theoretically a mathematician and could easily look up and understand information theory in detail, I can only conclude that he is being intellectually dishonest.