Simplicity has advantages over biological accuracy
Skarda and Freeman fail to show that their dynamic models have any real advantages. Connectionist except simplifications in order to get useful scientific results.
For example:
- they leave the role of chaos vague;
- they argue ineffectively that pattern completion differs from the activity of real networks;
- although it is true that feedback is important, there is no point in "blindly simulating neural circuitry".
Ronald Rosenfeld, David Touretzky, and the Boltszmann Research Group 1988.