MIT

Researchers have developed an analog silicon computer chip with about 400 transistors that mimics the activity of a single brain synapse - a connection between two neurons that allows information to flow from one to the other.

The day that computers outsmart their human overlords may yet lie in the distant future, but a new computer chip that mimics the basis of learning and memory in the brain is a critical step towards that moment.

"We are not talking about recreating a whole brain at this point. We have to start with one system," Chi-Sang Poon, a research scientist in the Harvard-MIT Division of Health Sciences and Technology, told me Wednesday. 

Poon and colleagues have started with an analog silicon chip outfitted with 400 transistors that emulates the activity of a brain synapse — a connection between two neurons that allows information to flow from one to the other.

There are about 100 billion neurons in the human brain, each of which has synapses — or gaps — between it and other neurons. Emulating one is a step "for building truly intelligent brain systems," he said.

The development is a departure from digital computer chips that simulate the spiking of neurons, treating their function like a simple on-off switch. Poon's chip directly mimics the ion channels that lead to the spiking. He likens it to understanding what's going on inside a black box.

"We really get into the nitty-gritty of how the neurons work intra-cellularly," he explained. "That involves all the ionic processes that are going on. Neuroscientists spend their life trying to understand how these things work and fit together."

The chip, he said, will allow neuroscientists to conduct basic research on how the brain actually works. Eventually, this could lead to the study and treatment of diseases related to brain malfunction, for example.

Other potential applications further down the road include devices that replicate specific brain functions that are incorporated with brain-machine interfaces. This could increase the versatility of devices that allow people to operate wheelchairs and computer mice with their thoughts, for example.

"Once it is to the level that we can build reasonably good replicas of brain systems, we can actually build brain systems that can replace some of the damaged brain parts," Poon added. 

The same concept, he noted, could be even be used to "enhance part of the brain systems beyond the normal human capacity."

And, stepping outside the machine-brain interface, the same chips could be used to build artificial intelligence devices that faithfully mimic or replicate brain behavior for tasks such as pattern recognition, cognition, learning, memory and even decision making.

"As long as we understand how the brain works, we can always reverse-engineer it and put it in a chip to reproduce those functions," Poon said. 

Poon and his colleagues describe the chip this week in the journal Proceedings of the National Academy of Sciences. 

More on brainy computer technology:

• IBM unveils brain-like chip
• IBM computer simulates cat's cerebral cortex
• Molecular computer mimics human brain
• Paralyzed man uses brain-powered robot arm to touch