Hadn't seen this Nanoge conference earlier this year. Just popped up in a search.
Not about Akida but is about our type of AI and is good to see what Huawei think of neuromorphic. Positive thoughts.
Though, they indicate in the abstract they are still playing with analog....maybe they'll wake up at some point
Trusting the other telecoms etc are starting to get their head around it and think the same way.
When I see Huawei and neuromorphic linked, I got back to the eX3 independent program that was using Akida in HiSilicon KunPreng 920s & what gets shared
Resources
www.ex3.simula.no
www.nanoge.org
Moraitis, Timoleon
Huawei Technologies
Making neuromorphic the main stream of AI
Authors
Timoleon Moraitis a, Qinghai Guo a, Hector Garcia Rodriguez a, b, Franz Scherr a, Adrien Journé a, Pontus Stenetorp b, Yansong Chua a, Dmitry Toichkin c
Affiliations
a, Huawei Technologies, Zürich, Switzerland, Zürich, CH
b, University College London, Roberts building, Torrington Place, London, GB
c, None, GB
Abstract
Neuromorphic computing is widely regarded as a constraint that increases AI efficiency but trades off proficiency, such as accuracy. However, this separates neuromorphic use cases from mainstream AI, into niches where proficiency is not crucial.
To the contrary, our recent work [1-5] shows that neuromorphic mechanisms do not constrain but rather they expand and improve mainstream AI in conventional measures of proficiency, while also improving efficiency. Specifically, we show that neuromorphic algorithms can outperform the state of the art of the broader, conventional machine learning field, beyond neuromorphic niches, in classification accuracy, robustness, inference speed, learning speed, and task reward. We show these in tasks such as keyword spotting, ImageNet classification, playing Atari games, controlling robots, many of which are usually out of reach for neuromorphic models. We test these in settings of online adaptation, supervised, unsupervised, self-supervised or reinforcement learning, meta-learning, and non-backprop-based deep learning.
This is achieved by exploiting also other biological mechanisms than spikes. The biological mechanisms that we employ include spikes, Hebbian plasticity, short-term plasticity, and efference copies. The algorithms remain suitable for efficient neuromorphic hardware, and include a new method for decreasing the power consumption of weighting operations in analog synapses.
Based on these results, we propose that exploiting the breadth of neuromorphic mechanisms, besides spikes, in suitable applications, is a path towards making neuromorphic the mainstream type of AI computing.