Skip to content ↓

Electronic circuit mimics the brain

Left to right: MIT researchers Richard Hahnloser, a postdoctoral fellow in the Department of Brain and Cognitive Sciences; H. Sebastian Seung, assistant professor of computational neuroscience; and Rahul Sarpeshkar, assistant professor of electrical engineering and computer science, were part of a team that created an electronic circuit that mimics the biological circuitry of the brain.
Caption:
Left to right: MIT researchers Richard Hahnloser, a postdoctoral fellow in the Department of Brain and Cognitive Sciences; H. Sebastian Seung, assistant professor of computational neuroscience; and Rahul Sarpeshkar, assistant professor of electrical engineering and computer science, were part of a team that created an electronic circuit that mimics the biological circuitry of the brain.
Credits:
Photo / Donna Coveney

Researchers at MIT and Lucent Technologies' Bell Labs report in the June 22 issue of Nature that they have created an electronic circuit that mimics the biological circuitry of the cerebral cortex, the brain's center of intelligence.

This latest advance in "neuromorphic" engineering -- creating devices that behave like neural systems -- was achieved by a team that included MIT researchers Richard Hahnloser, a postdoctoral fellow in the Department of Brain and Cognitive Sciences; Rahul Sarpeshkar, assistant professor of electrical engineering and computer science; and H. Sebastian Seung, assistant professor of computational neuroscience.

"Like electronic circuits, the neural circuits of the cortex contain many feedback loops," Professor Seung said. "But neuroscientists have found that cortical feedback seems to operate in a way that is unfamiliar to today's electronic designers. We set out to mimic this novel mode of operation in an unconventional electronic circuit."

The circuit was designed in collaboration with Rodney Douglas and the late Misha Mahowald from the Institute of Neuroinformatics in Switzerland. Much of the research was carried out at Lucent Technologies' Bell Laboratories in Murray Hill, NJ, where Professors Sarpeshkar and Seung are consultants.

In the future, general principles illustrated by this circuit could lead to hardware that efficiently accomplishes complex perceptual tasks such as recognizing objects by sight.

COMPETITION AND COOPERATION AMONG NEURONS

The circuit is composed of artificial neurons that communicate with each other via artificial synapses. All of these elements are made from transistors fabricated on a silicon integrated circuit.

Like neurons in the cortex, nearby artificial neurons affect each other. There also is an inhibitory neuron that receives input from the 16 excitatory neurons and returns inhibition to them. This inhibitory feedback keeps in check excitatory feedback that can lead to explosive instability.

In the brain, synaptic feedback connections are thought to mediate neurons' cooperative and competitive interactions. Such interactions are expressed most strongly in the circuit when multiple stimuli are presented at the same time.

When simultaneous electrical currents are applied to two artificial neurons, the circuit responds to only one stimulus and suppresses its response to the other, much like a frog choosing which of two flies to strike at.

Like the brain, there is no single element in the circuit that decides which stimulus to suppress. The decision is the outcome of an emergent, collective property of all the neurons.

DIGITAL AND ANALOG

A typical neuron in the brain might be connected to 10,000 other neurons. Because there are billions of neurons, this makes the brain a vast and intricate network. "Biologists like to focus on simple linear pathways through this network, ignoring the tangled web of feedback loops, which seem too complex to even contemplate," Professor Seung said. "But it seems unlikely that we could ever understand intelligence or consciousness without understanding the role of feedback in the neural networks of the brain."

Because electrical engineers rely heavily on feedback in their designs, researchers have been tempted to draw analogies between electronic and neural circuits. But recent neurophysiological experiments suggest that the brain does not use feedback in the same way as conventional electronics, which is distinctly either analog or digital.

Perception, the authors write, combines digital and analog aspects. When we see an object such as an approaching car, we also receive a continuous stream of information about its color, its changing size in relation to its distance from us, its spatial relations to other objects and so on. Nevertheless, the digital component is still there because regardless of how the object appears, our brains make an either-or decision: is it a car or not?

Professor Sarpeshkar recently suggested that the hybrid analog-digital nature of the brain may be very important for its computational efficiency. "The electronic world is evolving more and more towards mixed analog-digital computation as the brain has already done," he said. "However, the brain's mixed-signal circuits combine analog and digital functions in a much more intimate way than is done in the electronic world."

"Philosophers and psychologists have long been struck by the duality between analog and digital in perception," Professor Seung said. "They have further speculated about whether the computational operations underlying perception in the brain are analog or digital.

"Our research suggests that the two sides of this duality are not mutually exclusive: the brain's neural circuitry is actually a hybrid in which analog and digital coexist."

Professor Sarpeshkar, who does active research in hybrid electronic circuits, said that "hybrid electronics has the potential to revolutionize computing in the future because it combines the digital advantages of programmability, noise immunity and divide-and-conquer processing with the analog advantages of efficiency." He thinks that "the most immediate applications of such biologically inspired circuits are likely to be in sensory data processing, where the input is analog, and in prosthetic applications for the deaf and blind, where mimicking the biology is important."

This work was supported by the Swiss National Science Foundation SPP Program, Lucent Technologies and MIT.

A version of this article appeared in MIT Tech Talk on July 12, 2000.

Related Topics

More MIT News

Headshot of Catherine Wolfram

A delicate dance

Professor of applied economics Catherine Wolfram balances global energy demands and the pressing need for decarbonization.

Read full story