The society of mind: Marvin Minsky

✍ Scribed by George N. Reeke Jr

• Publisher: Elsevier Science
• Year: 1991
• Tongue: English
• Weight: 466 KB
• Volume: 48
• Category: Article
• ISSN: 0004-3702
• DOI: 10.1016/0004-3702(91)90034-h

✦ Synopsis

The Society of Mind (SOM) represents Marvin Minsky's brave attempt to reconcile traditional artificial intelligence (AI) with connectionist approaches that eschew explicit programming in favor of self-organizing systems of simple processors connected by weighted links. In SOM, Minsky deals, often implicitly rather than explicitly, with fundamental issues in the philosophy of mind that are at the heart of current debate on the theoretical possibility of achieving AI and on the relative merits of various approaches to AI. In this review, I shall try to indicate the nature of some of these issues, then give a few vignettes from SOM that illustrate Minsky's approach. Finally, I will point out some difficulties that this approach appears to share with conventional AI and suggest that careful attention to the biology of organisms that display adaptive behavior may be well rewarded.

Minsky is well aware of the pitfalls in seeking a reduction of intelligence to mechanism. He is careful to avoid the so-called "homunculus problem", which plagues most attempts to break down intelligent behavior into simple components. This problem arises when the output of a particular unit being studied turns out to require intelligent interpretation elsewhere in order to play its role in the overall behavior of a system. An infinite regress ensues in which intelligence can never be localized to any of the units studied. Minsky deals with this problem by assuming that tasks can be subdivided into simpler and simpler subtasks until the agents required to carry out the subtasks are * (Simon & Schuster, New York, 1986); 339 pages, $10.95 (paperback).