Fujitsu Ltd reports that it has simulated the operation of a neural computer with the processing capacity of 100,000 neurons or neural nodes, compared with, it says, only six neurons for current neural computers. It also says that its simulator processes data 10 to 100 times faster than the fastest simulators announced by US academic institutes. It says that it plans to build a prototype industrial robot applying a simplified version of neural circuitry by the middle of the year, and a prototype neural computer in two to three years. Neural computers mimic the working of vertebrate nervous systems, and their key attraction is that they are heuristic – have the ability to learn by application of successive approximations. Neural circuitry is under development at AT&T in the US, and TRW Inc has a neural network computer that is front-ended by a DEC VAX. A number of start-up companies have also been formed to pursue the technology.
FUJITSU REVEALS PLANS FOR NEURAL COMPUTER
