DARPA Building Neuromorphic Brain on a Chip (Paging Sarah Connor)

Illustration for article titled DARPA Building Neuromorphic Brain on a Chip (Paging Sarah Connor)

DARPA's brain-on-a-chip project (cleverly titled SyNAPSE, or Systems of Neuromorphic Adaptive Plastic Scalable Electronics) sounds straight out of Cyberdyne's portfolio: They want to "develop a brain inspired electronic 'chip' that mimics that function, size, and power consumption of a biological cortex." That whole neuromorphic adaptive business sounds a whole lot like the T-800's neural net processor, don't it? Here's the scary manifesto that puts us on the path to Judgment Day.

As compared to biological systems, today's intelligent machines are less efficient by a factor of one million to one billion in real world, complex environments. The key to achieving the vision of the SyNAPSE program will be an unprecedented multidisciplinary approach that coordinates aggressive technology development in the following technical areas: 1) Hardware; 2) Architecture; 3) Simulation; and 4) Environment. Hardware includes neuromorphic electronics with novel, high density, plastic, synaptic components; Architecture includes neuromorphic design from microcircuits to complete system; Simulation includes large-scale digital simulation of neuromorphic circuits and functional neuromorphic systems; and Environment includes virtual training, testing and benchmarking for neuromorphic systems.


We can only delay it, not stop it. [Danger Room]


That's interesting, because DARPA are already funding work in this area under the NeoVision program. I was working on this at my last job. The objectives sound similar, although in the first phase of NeoVision we were concentrating on duplicating the functionality of the early visual processing only.