“Spiking neural network that imitates the brain’s underlying neural representations and behavior.”
Intel has released an eight million neuron “neuromorphic” system to the wider research community, hoping to foster a new level of human-mimicking computer systems.
Operating under the codename Pohoiki Beach, the system is constructed using a series of Nahuku chips boards; each of which contains eight to 32 Intel Loihi neuromorphic chips. Neuromorphic is a branch of computer engineering that aims to imitate the neuro-biological architecture of the human nervous system.
Loihi enables users to process information up to 1,000 times faster and 10,000 times more efficiently than CPUs for specialized applications
The Loihi is Intel’s 60 mm2 chip fabricated using the 14nm fabrication process that Intel is getting under mass production. The Loihi combines a host of features such as dendritic compartments, hierarchical connectivity and synaptic delays, all of which help it mimic the human nervous system.
Chris Eliasmith, co-CEO of Applied Brain Research and professor at University of Waterloo commented in a release: “With the Loihi chip we’ve been able to demonstrate 109 times lower power consumption running a real-time deep learning benchmark compared to a GPU, and 5 times lower power consumption compared to specialized IoT inference hardware.”
“Even better, as we scale the network up by 50 times, Loihi maintains real-time performance results and uses only 30 percent more power, whereas the IoT hardware uses 500 percent more power and is no longer real-time.”
Pohoiki Beach System
The Loihi chip and Pohoiki Beach system is currently been used in projects around the world that require scalable spike-based computation system. At the Telluride Neuromorphic Cognition Engineering Workshop, in Colorado, US, researcher showed how the system can provide critical adaptation capabilities to artificial limbs like the AMPRO prosthetic leg.
Due to its computational spike efficiency it’s an ideal system for the coming onslaught of edge computing, much of which will require quick computational jumps like smart video surveillance systems that are only programmed to act during certain sensed events.
Professor Konstantinos Michmizos of Rutgers University commented that: “Loihi allowed us to realize a spiking neural network that imitates the brain’s underlying neural representations and behavior. The SLAM solution emerged as a property of the network’s structure. We benchmarked the Loihi-run network and found it to be equally accurate while consuming 100 times less energy than a widely used CPU-run SLAM method for mobile robots.”