Silicon Brains: Not as Cuddly as the Real Thing.
Are you real? What is ‘real’, more of a philosophy question than a scientific one, but what if a computer worked like your brain? What if, one day the line between computer and human were blurred? That day might be coming sooner than you think.
Currently there are two major problems with designing a robotic brain. The first is hardware, the brain is an incredibly complex thing that we don’t even fully understand, even if we could theoretically produce something close to that work of art there is that second problem– The software, designing software to take advantage of that type of power would take something short of genius to do, especially if it were going to be something easy enough that you or I could use.
While the second part of the problem still lingers, the first part is getting closer to reality. Stanford scientists have produced a circuit boards modeled after the human brain that are 9,000 times faster, while using significantly less power. This is just the start, the chips were produced using 15 year old fabrication technologies [yes, you read that correctly one five YEARS OLD!!], at the time costing roughly $40,000.
Of course, even with speeds like that it is still a slow poke compared to the brain. In comparison, the cortex of a mouse is roughly 9,000 times faster than the average computer, but the average computer isn’t just slower it takes 40,000 times more power to operate compared to this new technology.
How is this computer different than the one you reading on? Well for starters it has 16 chips on it, each called a Neurocore processor [again based on the way the brain is built]. All 16 chips together simulate roughly 1 million neurons and billions of synaptic connections– all of which is about the size of an Ipad and takes roughly the same amount of energy to run.
And the other problem, the software issue– the team is working on it:
“Right now, you have to know how the brain works to program one of these,” said Kwabena Boahen, associate professor of bioengineering at Stanford. “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.”
Lot’s of work still lays ahead, reducing costs by using newer fabrication methods and increasing production volume will bring the costs down to about $400 a board. The Neurogrid is about 100,000 times more energy efficient than a personal computer simulation of 1 million neurons, but is still a far cry from being as energy efficient as the human brain, or anywhere near as quick.
“The human brain, with 80,000 times more neurons than Neurogrid, consumes only three times as much power,” Boahen writes. “Achieving this level of energy efficiency while offering greater configurability and scale is the ultimate challenge neuromorphic engineers face.”
Know the difference between a hippopotamus and a hippocampus? You might be interested in the full article — here!
Benjamin B.V., Gao P., McQuinn E., Choudhary S., Chandrasekaran A.R., Bussat J.M., Alvarez-Icaza R., Arthur J.V., Merolla P.A. & Boahen K. & (2014). Neurogrid: A Mixed-Analog-Digital Multichip System for Large-Scale Neural Simulations, Proceedings of the IEEE, 102 (5) 699-716. DOI: 10.1109/JPROC.2014.2313565