Dembski's argument contains the concept of the 4th law of thermodynamics. He argues that for a 'closed system' CSI or complex specified information cannot increase. Although he does allow for a small increase he basically claims that
I(A&B) = I(A) Mod UCB
I(A) = - log(2)P(A)
I(A): Information in event A
P(A): Probability of event A
UCB: Universal complexity bound I(UCB)<500 bits or P(UCB)<= 10^-150
But there seem to be several examples which appear to contradict his claims.
First of all Tom Schneider's
Evolution of biological information shows how a simple mutation selection algorithm can increase the information in the genome.
Adami as well shows in
Evolution of biological complexity how selection/mutation acts like a Maxwell demon, increasing the information of the genome, but without violating any laws of thermodynamics.
In fact, Dembski's argument applies to closed systems (sounds familiar 2nd law fans?)