TOWARD THE DEVELOPMENT OF ARTIFICIAL STUPIDITY by John Koch, Madison, WI c 1989 Journal of Irreproducible results Current attempts to design machines with Artificial Intelligence (AI) present many technical and philosophical challenges. The complementary search for Artificial Stupidity (AS) presents even greater challenges, while holding out the promise of immediate and concrete social benefits. At first, this might seem paradoxical. But those of us involved in AS research believe that many of humanity's problems arise not from a lack of natural intelligence but an excess of natural stupidity. Intelligent people do foolish things that waste time, cause worthwhile projects to fail, and sometimes even produce death and injury. If artificial systems could take over responsibility for doing dumb things, a great deal of natural intelligence would be released. The conservation of the intelligence we already have is more cost-effective than the attempt to create more of it artificially. ............................... ............................... Our first task has been to establish an operational definition of stupidity. We believe that stupidity is in evidence when an organism takes active, self-generated measures that lead to the frustration of the goals which thos actions were meant to achieve. So far, objectively-verifiable stupidity seems to be a uniquely human trait. Though certain inanimate objects, such as rocks, computers, and breakfast cereals (for example, Cream of Wheat) are proverbially stupid, they are incapable of self-generated action. Animals may do things that appear to be stupid, but we know too little about their motivations to say that without a doubt, for example, my dog did not have a perfectly reasonable end in mind towards which digging up your garden was an effective means. We have decided to adopt an extension of the Turing test to objectively verify stupidity: we will know that we have Artificial Stupidity when a human interacting with a machine cannot tell the difference between it and a)his or her boss, or b)the stupidest member of his or her spouse's immediate family. While making only slow progress on the heuristics that would allow independent generation of dumb ideas by a machine, we have begun working on the "idiot" programs that will provide the misinformation base for those ideas. One fertile source has been television and movie scripts, particularly in the dialogue given to professionals sucha as police officers, doctors, social workers and scientists. We have made progress in the biological sciences by using the scripts of 1950's science fiction movies to create a program that responds either "I don't know what it is, but I've never seen anything likee it in my life," or "It appears to be some kind of insect venom," when presented with over 50 different substances, including shoelaces, cheddar cheese, a California Pinot Noir, flour, dried parsley and a copy of the New York Times Review of Books. We are anxious to develop similar programs in other disciplines, and would be greatful for assistance in creating a cross-disciplinary resource file of stupid people. Progress in AS research will probably be painstakingly slow, and its first fruits disappointingly primitive. After all, no computer, however powerful, can match the stupidity potential of the human brain, especially when several human brains are working together on a committee. If we are to develop a system that is capable of lifting the burden of making stupid mistakes from the shoulders of humanity, we must find a way to link many Artificial Stupidity Systems together.