FLUID CONCEPTS AND CREATIVE ANALOGIES: Computer Models of Mental Fluidity and Creativity

by Douglas R. & The Fluid Analogies Research Group Hofstadter ‧ RELEASE DATE: Feb. 28, 1995

Artificial intelligence expert Hofstadter (GÖdel, Escher, Bach, not reviewed, etc.) challenges conventional computer simulations of reasoning. These simulations don't begin to match the richness and flexibility of human thought, Hofstadter says: They're either ""brute force"" performances that simply take advantage of the computer's speed in considering already established options or programs that provide limited information that leads to a foregone conclusion. As an alternative, Hofstadter and his students create computer programs that model anagrams and analogies (remember those SAT questions: ""A is to B as C is to ...""?) as examples of human thinking and creativity. They create programs that allow the computer to search and discover candidates for the missing terms chosen from a ""coderack"" (Hofstadter loves puns). They also allow for ""slippage"" -- deviation from strict rules, which is what Hofstadter means by ""fluid concepts"" (for example, what is the solution to ""ABC is to XYZ as ABD is to ...?). Reading Hofstadter gives clues to how people -- and presumably his computer programs -- slip around these barriers to come up with answers that are described variously as ""happy,"" ""low temperature,"" or ""urged"" with certain ""pressures."" Hofstadter's admittedly complex writing style also has a wonderful colloquialism: You can hear him talking to his students, in part to get his own thoughts straight in the process. Reading this compendium of articles on games he and his AI researchers have programmed leads to consideration of human thought processes. In contrast to the programs of others in the field, Hofstadter's games are modest, played within small ""domains."" But they open up ideas on how perception and concept formation are linked in parallel processing tracks in the brain. For Hofstadter, the art of programming a computer is not an end in itself but a means to further understanding the mind at work. An excellent and updated review of a major trailblazer's spin on AI.

Pub Date: Feb. 28, 1995

ISBN: N/A

Page Count: 512

Publisher: Basic

Review Posted Online: N/A

Kirkus Reviews Issue: Dec. 15, 1994

Categories: NONFICTION