Initial, cursory thoughts on Dennett's Intuition Pumps...
Intrigued by his conception of Design Space, seems to have some possibilities for being fleshed out. Dennett is mostly about the intuitive and conceptual, hence the title, so it is in a (definitional) sense vague.
Dennett understands better than almost all the other materialists just how unlikely working designs are, and doesn't like that materialists shrink from calling well-engineered devices in biology "design." How he seems to reconcile it that he imagines the solution space in metazoan architectures to be well-connected and smoothly connected -- to have significant structure.
This is something that I am thinking of as Speculative Darwinian Recursion: If it is complicated, imagine it resulting from a plausible amount of tinkering on a slightly less complicated, slightly more plausible precursor. Dennett calls this "lifting" and, much as Darwin himself analogized natural selevtion from the deliberate process of artificial selection (i.e. breeding), analogizes with engineering efforts, with leveraging new technology form old.
In one of his hard AI arguments in the book he discusses a program for generating new compositions based on style samples from composers. Give it enough Bach and it will create more -- well, more Bach-esque material. Give it multiple composers' material and it will amalgamate the material. I don't know that he is arguing necessarily (although it seem implicit) that this application is a trivial program, but he seems to suggest that there is nothing significant about programming into the application an insight into musical style itself, as opposed to simply writing a program that had enough Bach insights "hardcoded" in to generate Bach-esque material.
This appears in a context of "creation without comprehension" -- which for Dennett applies to both AI and biology -- and follows the example of the chess program that was able to beat the human chess-player Gary Kasparov. Dennett asks us to ignore the brute-force approach of the application, as well as any path-pruning heuristics that were built into the program, and accept that in principle this is really not fundamentally different from comprehension and reasoning based on insight.
Are these intuition pumps or conceptual circularities?
Note: Revisit Roger Penrose's and Douglas S. Robertson's thoughts on insight.