Programs that are designed to imitate the way the brain works are called neural nets.

Although Frank Rosenblatt of Cornell University experimented with artificial intelligence using neural nets in the early 1960's (Yount, 1996, p. 50), interest in neural nets was not strong until the mid 1980's as computers became more powerful.

Scientest began to write neural net programs that strengthened and weakened connections gradually. When two connected "neurons" fired at the same time, their connection was strengthened. This is how some scientists think the brain works.

Learning to Talk

Terry Sejonowski of Johns Hopkins developed NETtalk, a neural net program with 18,000 synapses! This amazing program processed printed letters and learned to produce the appropriate speech sounds. It started out just making nonsense sounds, but "literally overnight, the program taught itself to say English words." (Yount, 1996,p. 52) "We found consistantlyy that a certain neuron turned on whenever the letter was a vowel. We hadn't taught the system about vowels or assigned that neuron to track them. That was somethign that came out of the system on its own." (Sejonowski, as quoted in Yount, 1996, p. 52)

In 1993, Intel put a neural net chip on the market, with 64 neurons and 8,192 possible connections, or "synapse." (Yount, 1996, p. 53)