In a context of European and American efforts aimed at simulating the brain, a Standford team has developed a circuit board called Neurogrid that models brain functions at up to a million neurons and a billion synapses. No larger than a tablet, Neurogrid is still far from the 80 billion neurons in the human brain, but it compares favorably to other, similar projects like IBM’s SyNAPSE (256 neurons and 1024 synapses) or the University of Heidelberg’s HICANN (512 neurons and 224 synaptic circuits for HICANN). Even more interesting, Neurogrid consumes 40,000 times less energy than a PC, according to Kwabena Boahen, Associate Professor of Bioengineering at Stanford, meaning that scaling it up shouldn’t be, at least power-wise, too much of a problem.

Although still in the prototype stage, Neurogrid paves the way for therapeutic use, especially for simulating movement for people with localized cerebral paralysis. Built based on sixteen “neurocore” processors and using years-old technologies, its cost currently exceeds $40,000. But according to K. Boahen, this could be reduced a hundred-fold with modern technology and mass production. The next step is to develop a neuro-compiler dedicated to the abstraction of neurons and synapses, and sensors that can read brain signals. Still a way to go…