Artificial Intelligence and Natural Folly
22 Mar 1995 17:30Natural minds evolved for specific purposes - survival under constraints of finite knowledge, intelligence, and time to act. Thus necessity of selection of experience and courses of action, imposition of order on sensory data, all implying needs, wants, drives. Thus, folly: errors (since perfection does not evolve), and more particularly emotional biases, superstitions, idées fixes. (Cite Dennett, James, Jahoda.) Thus, natural folly is probably unavoidable. Artificial intelligence: interest and selection are necessary features of finite intelligence (though not of course sufficient). We must suppose therefore that AIs --- or IAs --- will exhibit a similar sort of "natural" folly. Mention, e.g., stereotyping by neural networks. (Observed, personally, in cognitive science 1.)
The Sphex as the type of natural folly, finitude dept.
History of natural folly. Castigation of folly is ancient and, while purely moral, startlingly ineffective. Not even Erasmus achieved anything in that direction. Practical remedies - memorization (rhapsodes), logic, argument (Sophists, Socrates, Plato). Experiment and skepticism. Diagnostic studies (e.g. Hume) early identified greed and other moral weaknesses, - a natural transition from moral stricture. See also: Artificial intelligence. Cognitive Science. Daniel Dennett. Imagination. Intellectual immune systems. Intellectual standards and competence. Intelligence augmentation. Judgment and Decision-Making. Memes. Memories. Neuroscience. Supersitition.
