We’re beginning to focus on neuromorphic scaling. Pohoiki Beach, with 64 Loihi chips operating as fast as 6us per timestep, is about 160x faster than any other neuromorphic system at this scale. Our partners are beginning to demonstrate applications at compelling power and performance levels, like SLAM, adaptive control of robot arms and prosthetics, audio processing, adversarial-robust inference, and one-shot olfactory learning. Looking forward to new breakthroughs enabled at scale! Chris Eliasmith Konstantinos Michmizos Thomas Cleland Franz Scherr Edward Kim #loihi #neuromorphic
Keith Rebello
5 年
Exciting! 64 chips is just a stepping stone. Looking forward to seeing how it scales!
Paul Sisneros
5 年
What is one-shot olfactory learning?
David Wyatt
5 年
Very Interesting ...
Leon P. Stavrou
5 年
Thanks Professor Themis Prodromakis for sharing Mike’s post.
Bryn Balcombe
5 年
Mike Davies it’s great to see Intel Corporation committing to Neuromorphic Computing. The approach offers significant benefits for high-speed robotics applications like Roborace in the domains of SLAM, path planning, audio processing and event based cameras. I look forward to seeing the innovations and breakthroughs Pohoiki Beach & Pohoiki Springs enable! ????
This is great! Choosing an algorithm to run with 8 million neuron and optimizing it will be a huge task!
Mike Davies
5 年
See also IEEE Spectrum coverage... (apologies neuromorphs, for the nerd quote!)
Hardware Engineering Leader - Delivering delightful products and experiences!
5 年That is a lot of chips on one board!? Would love to see how this can get into portable solutions.??