Communications of the ACM
Blueprint For an Artificial Brain
The Bielefeld memristor built into this chip is 600 times thinner than a human hair.
Credit: Bielefeld University
Bielefeld University's Andy Thomas is developing an artificial brain using memristors, which are electronic microcomponents that imitate natural nerves. Having already developed a memristor capable of learning, Thomas and his colleagues are now trying to put the memristors together in a brain blueprint.
Similar to an electronic version of a synapse, a memristor connects electric circuits and learns from previous impulses sent by the circuits. Thomas says the similarity to synapses makes memristors especially fitting for creating an artificial brain and a new generation of computers.
To make a neuromorphic computer work, certain principles of nature must be transferred to technological systems. For example, memristors must recall earlier impulses and neurons must only respond to an impulse that goes beyond a specific threshold.
Learning responses, such as those demonstrated in Pavlov's classic dog experiment, are possible with memristors because they are capable of greater information storing refinement than the bits on which previous computer processors have been based, Thomas says. For example, while bits only have on and off modes, memristors can raise or lower resistance continuously.
"This is how memristors deliver a basis for the gradual learning and forgetting of an artificial brain," Thomas notes.
From Bielefeld University
View Full Article
Abstracts Copyright © 2013 Information Inc., Bethesda, Maryland, USA
No entries found