Worth a read if you're a Douglas Hofstadter or Karl Friston fan.

Significant bit from the conclusion: "The theory is testable and falsifiable at the level of computational modeling, but also in terms of neural implementation."