The brick wall: NP completeness

For over 40 years, artificial intelligence (AI) has not fulfilled its promise of truly intelligent machines for general use. As early as the 1950s and 1960s, scientists developed computational models of intelligence. They then excitedly coded these models into the best computers of the day. At first the scientists were puzzled by the machines' inability to produce reasoned output. Bewilderment became frustration when they realized they had banged into an unforeseen brick wall. This wall had stopped them in their tracks and continues to do so today. AI also has stymied scientists and engineers in other fields such as operations research (the field concerned with determining efficient manufacturing and scheduling protocols), VLSI chip design and testing, and database management. The brick wall exists because many combinatorial problems that are fundamentally important are NP complete. (NP stands for nondeterministic polynomial time.) We illustrate the problem NP completeness causes with an example taken from manufacturing. A general purpose welding device is to be used on an assembly line. It will make numerous welds at predetermined positions on a particular kind of part. The positions will be welded in a specific sequence called a schedule. This schedule is programmed into the welder before a "run".