When designing a machine that makes decisions and judgment calls, as is the essence of A.I., there are two different paths to take: Strong A.I., and Weak A.I.. Besides the obvious literal interpretation that Strong A.I. may be more capable than Weak A.I. (which isn’t true), there are more formal meanings to the terms Weak A.I. and Strong A.I.:

Weak A.I. is any artificial intelligence (usually in the form of a computer program) which does not aim to operate in the same way as, or resemble, human thought. Rather, Weak A.I. is generally an engineering approach to artificial intelligence, a means to an end.

Strong A.I. is, on the other hand, a form of artificial intelligence which aims to “think” in the same way the human mind does, and even to eventually out-perform it. Strong A.I. mimics the cognitive processes of the human brain.

Hofstadter was upset with one of his own works (a computer program) after submitting it to a competitive assignment he designed while teaching his first A.I. class. The goal was to extrapolate data from number sequences (similar to the kind I mentioned here). He crammed a lot of knowledge into his program in the form of rule bases (heuristics) and let it loose performing a massive amount of calculations. What upset him about this was that he lost sight of his original research goals. He wanted to mimic the way the human mind thought, and this wasn’t the way.

The human mind doesn’t brute force problems by performing thousands of calculations a second like a computer can. Rather, it is really fascinating (and often difficult) to break down and analyze the way humans can think problems through. This is the goal of Strong A.I., and the goal Hofstadter set for himself: to design computer programs which steps through the same processes as human thought, and to learn something about cognition in the process.

-James