Read the essay From Micro-Worlds to Knowledge Representation: AI at an Impasse, by Hubert Dreyfus, which is Chapter 6 of Mind Design II.
This article examines some well-known symbolic AI projects from the 1970s. Dreyfus argues that using "microworlds" to represent some small portion of the real world in isolation is fundamentally flawed, because in order to genuinely understand something in a microworld setting requires understanding a much larger matrix of related concepts. He seems to be saying that any attempt to model some sub-part of the world apart from the whole is doomed to failure, because any sub-part has countless indispensible connections to the larger world which, if omitted from the model, would render it meaningless. Do you agree? Is it possible in principle to isolate some subset of knowledge from the larger universe of all knowledge?
Dreyfus also argues against the claim of symbolic AI that both thinking and perception can be explained in terms of formal rule-governed symbol-manipulation processes. To him, many aspects of intelligence are simply not formal processes. For example, could the complexity and subtlety of emotional responses, intuition, or creativity be captured by a set of rules in some sort of formal descriptive language, however sophisticated? Dreyfus states on page 179: "I will attempt to lay out the argument which underlies my antiformalist, and therefore, antimechanist convictions". This seems to imply that he sees these terms as equivalent. Could a system be formal on one level (that is, governed by a set of precise rules), but informal on another? Is it possible to be an "antiformalist" while still believing in "mechanism"?
You don't have to turn in anything for this assignment. Just think about the above questions, and be prepared to answer a few brief questions about the reading when you come to class.