Brain-like efficient (20W) AI

Watching Prof. Thomas Nowotny on building more energy-efficient, brain-like AI:

https://youtu.be/p0wELKagiLY

The idea is to get the benefits of AI with something more like 20W the human brain uses. To do that, you move away from current techniques to “neuromorphic computing”, which is more brain-like inspiration in technology.

The video features spiking neural networks and breakthrough in training these things via “loss shaping”.

I’d not heard of Intel’s Loihi research chip:

It incorporated self-learning capabilities, novel neuron models, asynchronous spike-based communication, and many other properties inspired from neuroscience modeling, with leading silicon integration scale and circuit speeds 

Sounds fun, and no you can’t just go out and buy them.  Looking at these slides there are a bunch of platforms for this kind of computing: SpiNNaker, BrainScalesS, Synsense, IBM synapse.