This Brain-Inspired Microchip Is 9000 Times Faster Than A Normal PC

This Brain-Inspired Microchip Is 9,000 Times Faster Than a Normal PC

You're looking at Neurogrid: a slab of silicon inspired by the human brain, which is 9000 times faster than a normal computer brain simulator and uses way less energy to boot.

Developed by a team of Stanford bioengineers, it's worth pointing out that this is hardly the first microchip to be inspired by the human brain -- they have come and gone in the past. It is, however, capable of simulating 100s and 1000s more neurons than any in the past, and on less power than it takes to run an iPad. The research appears in an article for the Proceedings of the IEEE.

It's no wonder scientists want to recreate the brain in silicon: even a mouse cortex can operate 9000 times faster than a PC, and a even then the computer uses 40,000 times the power, too. Hence Neurorgrid, which uses 16 custom-designed "Neurocore" chips to simulate one million neurons and billions of synaptic connections. It's 9000 times faster and 100,000 times more energy efficient, than a personal computer simulation of one million neurons.

Understandable, then, that the prototype's worth a cool $US40,000 -- but sadly it's not easy to code for. Kwabena Boahen, one of the researchers, explains:

"Right now, you have to know how the brain works to program one of these. We want to create a neurocompiler so that you would not need to know anything about synapses and neurons to able to use one of these."

Indeed, the idea is to use the devices to control prosthetic limbs for paralysed people -- a fitting application for a synthetic brain. But there's still plenty of potential to make this thing way more cheaper -- and, in turn, way more practical for the real world, too.

Currently, the 16 Neurocores, each supporting 65,536 neurons, are made using 15-year-old fabrication technology. Switching to modern production techniques could slash the cost by a factor of 100, making a million-neuron board just $US400. And then they could be used to control just about anything. [Proceedings of the IEEE via Stanford]