Following the announcement last September of the 8 million neuronal neuromorphic system, codenamed Pohoiki Beach, Intel today announced the expansion of its neuromorphic system to 100 million neurons, the equivalent of the number of neurons in a small mammalian brain.
It’s easier to understand, with a ladybug brain with about 250,000 to 500,000 neurons, a cockroach brain with about 1 million neurons, and a zebrafish brain with about 10 million neurons.
Of course, the Pohoiki Springs system is still based on Intel’s neuromorphic processor Loihi. Like the brain, Loihi can handle demanding workloads at 1,000 times faster than conventional processors and 10,000 times more efficiently. Pohoiki Springs is the next step in extending the Loihi architecture and can be used to assess its potential to solve AI problems and a range of computational challenges. Intel researchers believe that neuromorphic systems have super-parallel and asynchronous signal transmission capabilities compared to today’s most advanced conventional computers, which can significantly improve performance while significantly reducing power consumption.
Pohoiki Springs, Intel’s largest neural-morphon computing system to date, uses a data center rack system that integrates 768 Loihi neuromorphic research chips into five standard server-sized chassis.
Intel will provide this cloud-based system to members of the Intel Neuromorphic Research Community (INRC) to extend its neuromorphic work to solve larger and more complex problems.
Data Center Rack System Pohoiki Springs (Source: Tim Herman/Intel)
“Pohoiki Springs has extended our Loihi Neuromorphic Research Chip by more than 750 times and is operating at less than 500 watts,” said Mike Davies, director of Intel’s Neuromorphic Computing Laboratory. Currently, some workloads run slowly on traditional architectures, including high-performance computing (HPC) systems. The Pohoiki Springs system allows our research partners to explore ways to accelerate the processing of these workloads. “
Intel and INRC researchers demonstrated Loihi’s abilities, including real-time recognition of gestures, reading Braille using new artificial skin, orientation using acquired visual landmarks, and learning new smell patterns. All of these functions consume only tens of milliwatts of electricity. So far, these small examples have shown excellent scalability, and Loihi is faster and more efficient than traditional solutions when running larger-scale problems. This mimics the scalability of the natural world, from the insect brain to the human brain.
It is important to note that neural morphing systems such as Intel’s Pohoiki Springs are still in the research phase and are not designed to replace traditional computing systems, but rather to provide researchers with a tool to develop and characterize new neuro-inspired algorithms for real-time processing, problem solving, adaptation, and learning. INRC members will use the Intel Nx SDK and community-contributed software components to build applications on Pohoiki Springs through cloud access.
Examples of promising and highly scalable algorithms currently being developed for Loihi include constraint gratification, search maps and patterns, optimization issues.