Neural networks are not limited to institutions like Dartmouth. Outside academia, the brainlike computer systems are being put to uses that non-neural programs can't match. A few examples:
• Airlines are beginning to use neural nets to allocate seats on flights.
• Security firms have devised systems that recognize individuals' handwriting or faces.
• Banks that have used neural nets to evaluate loan applications report fewer defaults.
• Neural networked electronic eyes look for parts defects in assembly lines.
• The Postal Service hopes one day to use the technology to read handwritten zip codes.
• The military has spent millions of dollars in hopes of revolutionizing target recognition, smart missiles, and self-navigating tanks.
And that's not all. Neural nets have also found their way into bond trading, oil drilling, linguistics, medicine, and the manufacture of products ranging from cars to fluorescent light bulbs. You can even try a neural network at home. A $99 program called Brainmaker (California Scientific Software, Sierra Madre) is made to be run on personal computers. Another PC neural net, called NeuroShell, can be had for $199 from Ward Systems Group in Frederick, Maryland.
But don't expect high-I.Q. computing for your money. Even the most advanced neural nets have a way to go before they become a match for human brains. According to the Pentagon's Defense Advanced Research Projects Agency, five years from now neural networks will be many times more sophisticated than today's versions but will only approach the complexity of a bee's nervous system.