Neuromorphic Computing at Cloud Level

AI is having an increasingly large impact on our daily lives. However, current AI hardware and algorithms are still only partially inspired by the major blueprint for AI, i.e. the human brain. In particular, even the best AI hardware is still far away from the 20W power consumption, the low latency and the unprecedented large scale, high-throughput processing offered by the human brain.

In this talk, I will describe our bio-inspired AI hardware, in particular our award-winning SpiNNaker2 system, which achieves a unique fusion of GPU, CPU, neuromorphic and probabilistic components. It takes inspiration from biology not just at the single-neuron level like current neuromorphic chips, but throughout all architectural levels.

On the algorithm front, I will give examples on how to use general neurobiological computing principles (hierarchy, asynchronity, dynamic sparsity and distance-dependent topologies/hierarchical computing) to reframe conventional AI algorithms, usually achieving an order of magnitude improvement in energy-delay product, for both inference and training.