[Model-Directed Learning of Production Rules,Bruce G. Buchanan and Tom M. Mitchell, p.297-312]

p.297 Knowledge-based artificial intelligence programs derive their power from the richness and depth of their knowledge bases. It follows that careful construction of the knowledge bases is an obvious prerequisite for high performance in such systems, yet we have few alternatives to hand-crafting these for each new program.

[Artificial Knowledge Systems, Barbara Hayes-Roth, p.333-346]

p.345 it has proven extremely difficult to discover any mechanism capable of producing anything approaching the intelligent and adaptive behavior typical of an average human being.

p.346 In the long run, the most promising avenue of approach toward the development of artificial knowledge systems may be to investigate and model human cognition as closely as possible.

[Rule-Based Understanding of Signals,H. Penny Nii, Edward A. Feigenbaum, p.483-501]

p.485-486 Each source of knowledge holds a considerable body of specialized information that a human expert would generally consider "ordinary." ...Also represented are the heuristics, that is, "rules of good guessing" a human expert develops in his area of expertise.