Researchers have developed an electronic circuit that mimics the wiring of the human brain. This could make for better speech and object recognition in computers, they say. And the rest of it…The research is not new, but in fact the culmination of more than two decades of work using transistors and silicon to simulate the natural circuitry in the brain. It is, however, the first time a circuit inspired by the brain’s cortex has been created in hardware. The circuit, built on a silicon chip the size of a fingernail, can’t learn the way the brain can, but its structure is based on one aspect of the brain’s functioning. A bit of basic neuroscience: neurons (in the brain) have specific thresholds. If an incoming signal does not meet this threshold, then a neuron will not fire. If signal is above threshold, then it will. If a signal is considrably above its threshold, the neuron will fire repeatedly, responding to the frequency of the train of pulses hitting it. The new circuit produces a simluation of connections between neurons with different weights, and the process of of local feedback between them. The chip, which works on the essential idea that neurons can be ‘on’ or ‘off’ in response to one another, shows the same filtering phenomenon that occurs in a network of neurons.And now, the tech. The chip was was constructed out of ordinary transistors arranged into a ring-shaped network of 16 artificial neurons and the synaptic connections between them. Each neuron connects to four of its neighbors, and like their biological counterparts, tends to fire when receiving incoming impulses. But the neurons also are connected to a central inhibitory neuron, which acts as a regulator. The regulator keeps the circuit from working itself into an uncontrolled chain reaction of neuronal firings, the researchers said. ‘This is a demonstration of what’s possible when circuits compute in biological ways,’ said Rahul Sarpeshkar, a computer science professor at the Massachusetts Institute of Technology who worked on the project. ‘But we’re still far away from building a brain.‘Researchers applied simultaneous electrical currents to two artificial neurons in their circuit. It selected the stronger of the stimuli and suppressed its response to the weaker. That is not unlike, say, a frog choosing which of two flies to eat, the researchers said. And the circuit, like a brain, maintained its selection as the weaker current was increased and the stronger decreased. This negative feedback allows neurons to amplify or suppress their firings according to the behavior of their neighbors, an analog trait that until now has been seen only in biological computers such as the brain. Whereas computers are purely digital, the chip’s creators argue, because they perform calculations in the binary logic of ones and zeros, this ‘biological’ computer blends digital and analog processing. Back to brain, if you will, where the way nerve cells fire in response to impulses received from their neighbors - responding to the frequency of the pulses they receive - looks very much like ‘a digital response’. But frequently, modulating the digital behaviour are factors such as delay between a pair of input signals - an analog process that couldn’t be readily mimicked by digital circuits. Until now. ‘It’s a hybrid of digital and analog,’ Sarpeshkar said. ‘In one simultaneous circuit, both digital selection and analog amplification can now co-exist.‘More on the MIT research behind the chip: web.mit.edu/newsoffice/nr/2000/machinebrain.html

The juice from Bell Labs: www.bell-labs.com/news/1999/october/25/2.html