Researchers in Stanford’s Brains in Silicon Lab, led by Associate Professor of Bioengineering Kwabena Boahen, have developed a system of chips that mimic the computing structure and process of the human brain.
The chips, which are custom-designed by the lab, are called “Neurocores.” They simulate over 6,000 individual neurons and many more synaptic, or inter-neural, connections and were designed with the aim of increased power efficiency as compared to existing chips.
The researchers combined the chips, whose development was funded by the National Institutes of Health (NIH), in an interlinked grid of 16 chips (called “Neurogrid”), which aims to mimic the human mind by simulating a network of one million neurons and several billion synaptic connections.
The design, which permits certain synapses to share hardware circuits, allows the Neurogrid device to operate with the same computing power and be the same size as an Apple iPad. However, the design also requires knowledge of the brain to effectively leverage the neural network. In the future, the lab aims to create a neurocompiler to compile computer code so that the network can run it more efficiently.