THE COGNITIVE COMPUTER: On Language, Learning, and Artificial Intelligence by Roger C. with Peter G. Childers Schank

THE COGNITIVE COMPUTER: On Language, Learning, and Artificial Intelligence

Email this review


The real excitement of Al comes when we realize what we have left out of our previous attempt, which then tells us what to try next."" So writes Roger Schank, chairman of Yale's Department of Computer Science, in this state-of-the-art treatise. Schank belongs to a theory-oriented school of artificial intelligence, as opposed to the product-oriented (e.g., makers of chess-playing machines, or robots). For him, and his talented grad students (many are the authors of the experimental programs described in the text), success in programming ""intelligent"" machines will come when we have a better theoretical understanding of how human beings use language, how they think and learn, build dynamic memories, make plans and set goals. Learning by analysis of human cognitive processes, and by error, the group has constructed programs of language understanding. Their aim is for machines to comprehend ordinary spoken English (not the limited words of BASIC or other computer languages). Over the years they have progressed from simple ""parsers"" that analyze sentences in terms of subjects and predicates, to analysis-in-context, generating programs geared to certain sets of relationships; they have devised sets of scripts (something like foreign phrase books with their pages on ""in the restaurant,"" ""shopping""), eventually designing programs that could make inferences, answer questions, paraphrase and translate--and sometimes achieving these outputs in one complex program. Schank sees these efforts as moving AI from a level of ""making sense"" of the input towards ""cognitive understanding."" No machine, he freely states, will ever achieve a level of truly human ""empathy."" The book begins with a strong, perhaps overlong, indictment of the present emphasis on the need for ""computer literacy."" It's the machines that have to understand us, not we them ! (We don't need automobile literacy to drive.) He ends with speculation on social change, questions of privacy and job displacement, along with some earnest thoughts on the proper use of AI. Some of his rosier thoughts on the democratizing capability of machines (access to education for all), as well as his boost for computer science departments to become entrepreneurial, will surely arouse controversy. The theoretical approach is also open to question but here, in the heart of the book, the careful development of ideas and the interesting examples score: one feels that progress in AI is slow but steady and real.

Pub Date: Nov. 5th, 1984
Publisher: Addison-Wesley