Neurogrid Simulates The Computing Power Of The Human Brain
This article is more than 2 years old
While many people, including Stephen Hawking, worry about the consequences of artificial intelligence vastly surpassing human intelligence, it makes sense to remind ourselves how much of our own computing power remains untapped. Understanding the way the brain computes is also helpful in understanding efficiency; rather than directing all the information through a CPU, which causes a bottleneck, the brain uses a vast network of neurons and synapses. So why not make the best of both worlds? Stanford University scientists have done just that, creating a circuit board called the Neurogrid, which imitates human brain function.
The Neurogrid has 16 interconnected “Neurocore” chips which allows it to “simulates a million neurons connected by billions of synapses in real-time, rivaling a supercomputer while consuming a 100,000 times less energy—five watts instead of a megawatt!” That’s roughly 9,000 times faster than a conventional computer. Stanford bioengineering professor Kwabena Boahen has been working on such a chip for years and chose the brain as a model because “from a pure energy perspective, the brain is hard to match.”
At $40,000, the Neurogrid is actually one of the cheaper brain-simulating developments, but scientists hope to reduce that price substantially, down to as little as $400. Such a chip could facilitate great strides in computing and robotics, as well as prosthetic limbs. Eventually, Boahen hopes that these chips will be able to control artificial apendages by translating brain signals into movements without the use of an EEG or power source. Instead, the chip would go directly into the brain of the patient. Right now, someone has to understand the human brain to program the Neurogrid, but researchers hope to create another component called a “neurocompiler” that would enable scientists and engineers to program and use the device without expertise in neuroscience.
The Neurogrid is one of many smaller projects designed to help understand the human brain, including the Human Brain Project, through which scientists hope to use a supercomputer to simulate human grey matter. But that’s much easier said than done. The human brain still has 80,000 times the neurons of Neurogrid, but only consumes three times the power. Altogether, our brains control our bodies on roughly 20 W of power, a staggeringly low number for all the brain does. As impressive as that is, we should probably enjoy our brains’ elite status as much as possible while we still can.
