Customer Reviews

Dynamic Memory Revisited

5 star:		(0)
4 star:		(1)
3 star:		(2)
2 star:		(0)
1 star:		(0)

 

 

In a style typical of Schank's writings, this book contains a number of grandiose claims about human memory without much support. The first two-thirds of this volume present a review of the much of Schank's previous work. The title is Dynamic Memory Revisited after all! The last third of the book, however, presents two new themes: nonconscious knowledge and goal-based scenarios for education. While he has extensively discussed goal-based scenarios elsewhere, the ideas relating to nonconscious knowledge appear to be new. Unfortunately, the exciting ideas associated with nonconscious knowledge and the "racing mind" reads more as a call for research than a presentation of scientific findings.

While Dynamic Memory Revisited provides a good overview of much of Schank's work, the poor quality of the writing tends to frequently get in the way. Schank's frequent restatement of ideas reads as a random narrative rather than an updating of important work. This book is worth reading for its ideas, but any reader must be willing to ignore presentation and detailed content in favor of high-level ideas.

If you are looking for strict information on AI, this is the wrong book... it talks about learning in human memory in terms of dynamic (always changing) memory structures. There is adequate discussion of scripts, MOPs and TOPS but not enough to model AI on.
While the writing style is a little dense, on reflection it's raw-ness helps to illustrate some of what Schank is saying. He updates ideas throughout the book, building to new information in the last chapters on case-based reasoning and non-concious knowledge. This is basically how learning and memory work: the mind builds on what it knows to synthesize new structures.

Schank is responsible for a number of good ideas in Artificial Intelligence/Cognitive Science; conceptual dependency, scripts, and case-based reasoning among others. I read the present book in order to find any new gems Schank might have dug up. While I believe that you can mine Schank's work for good ideas there are some shortcomings as well. There is no complete theory presented. You could not write AI code based on this book. Furthermore, I would want such a theory and would have to test it before I would accept any of Schank's suggestions for how education should be changed. I find many of Schank's ideas about education faddist. But then much of the current discussion on education is nonsense. School is work. Period. If you don't like work you will not like school. So what.