The Logic Theorist and its children: AI in action

In 1956, Allen Newell, J. C. Shaw, and Herbert created a program called the Logic Theorist, arguably the first program to show novel behavior and thus count as "artificial intelligence." The program was pitted against chapter two of Principia Mathematica, Bertrand Russell and Alfred North Whitehead's attempt to systematize and codify the principles of pure mathematical logic. The program succeeded in proving thirty-eight of the first fifty-two theorems presented there, but much more importantly, the program found a proof for one theorem which was more elegant than the one provided by Russell and Whitehead. Newell, Shaw, and Simon attempted to publish their results in a short article, listing the Logic Theorist as a co-author, but the Journal of Symbolic Logic was evidently unwilling to publish a paper cowritten by a machine. Not discouraged, Newell and Simon reported in their 1958 paper, Heuristic Problem Solving: The Next Advance in Operations Research:

There are now in the world machines that think, that learn and that create. Moreover, their ability to do these things is going to increase rapidly untilčin a visible futurečthe range of problems they can handle will be coextensive with the range to which the human mind has been applied. (Kurzweil 69)

Newell and Simon were, for a while, a bit of an embarrassment to the field of AI because of their optimistic predictions. The 1958 paper claimed that in ten years a digital computer would be the world class champion, and Simon predicted in the late 1960's that by 1985 "machines will be capable of doing any work that a man can do" (Kurzweil 69). Despite this overoptimistic futurism, Newell, Shaw, and Simon had laid the seeds for research in AI by showing both that a digital computer could parse and extract complicated logical forms and that a digital computer could produce novel behavior, also known as emergent behavior, cognitive science's term for behavior more complicated than expected. This focus on emergent behavior and reducing areas of human intelligence to formal logic systems led directly to most of AI research for the next fifteen to twenty years.

Some of the more well-known AI projects that followed included: Student, by Daniel G. Bobrow, which could solve algebra word problems and reportedly did well on high school mach tests, Analogy, by Thomas G. Evans, which solved IQ-test geometric analogy problems, and Terry Winograd's SHRDLU, which, as Ray Kurzweil drily puts it, "could understand any meaningful English sentence, so long as you talked about colored blocks" (Kurzweil 70).