Compared to even the fastest supercomputers on the planet, the human brain wins out at speed and efficiency. For the 100 billion neurons firing, the brain only uses just 20 watts of energy. A supercomputer uses 1 million watts of energy.
Now scientists at Stanford University have developed a computer chip that can simulate 1 million neurons and billions of synaptic connections in real time. Such a chip could drive robotic, prosthetic limbs with the same speed and efficiency of biological arms and legs.
The team, lead by Kwabena Boahen, associate professor of bioengineering at Stanford, created their so-called Neurogrid with efficiency in mind. It’s comprised of 16 computers chips, called neurocores. Instead of building the transistors on the chip to correspond with digital logic, they built them to correspond with the voltages present in brain cells.
Additionally, the amount of current corresponds with synaptic input. All of this saves power, says Boahen. In all, the Neurogrid is 100,000 times more energy-efficient than a standard PC attempting to simulate 1 million nuerons.
But even though Neurochip is more efficient, it still consumes more power than our brain.
“The human brain, with 80,000 times more neurons than Neurogrid, consumes only three times as much power,” Boahen wrote in an article for the Proceedings of the IEEE. “Achieving this level of energy efficiency while offering greater configurability and scale is the ultimate challenge neuromorphic engineers face.”