IBM has just announced that it will head up a collaboration among five universities and its own researchers to develop a computer that will mimic the function of a mammalian brain. But isn’t that what neural network technology does? Not quite.
In the past couple of years neural networks have come into wide–almost everyday–use in science and medicine, industry and business. They have been hailed for their ability to solve complex problems requiring prediction and classification, and they have been used for decision evaluation, production forecasting, and other questions that arise in operations management; in medicine, where their capacity for classification has become valuable, they now play a big role in predictive diagnostics.
The way neural networks manage to accomplish all this cool stuff is by mimicking the activity of neurons in the biological brain. But the lead IBM researcher points out that the way a brain actually works is through a network of synapses, the chemical junctions through which neurons communicate with other neurons and with cells.
The difference between the two types of networks may sound too subtle to be important, but for a computer, it’s a tall order. The project is also a cat-and-mouse game that will involve scientists and scholars from fields as diverse as psychology and materials science in addition to computer scientists. Some of these experts have participated in a recent attempt to simulate the synaptic networks in a mouse brain, and the new project’s ultimate goal is to create electronic circuits that simulate the workings of a cat’s brain. They’ve got the mouse, so as those famous mice Tom and Jerry love to say, bring on the cat!