Embedded AI is gaining traction for reasons of privacy, latency, safety of operation and energy consumption. Dedicated hardware accelerator must therefore be designed and fabricated, along with the associated learning and quantization strategies. The main challenge to be solved is the energy dissipation, and in data centric applications such as AI, the main source of energy comes from moving data. Current accelerators try and leverage weight quantization as well as sparsity of activations. We will get insights into what semiconductor technology can bring in that respect.

However, when compared to what biology achieves, the current state-of-the-art is still orders of magnitude away in terms of energy efficiency. We will see how brain inspiration does translate into circuit or technology specification. We will explore what spiking neural networks can bring, exploiting novel technologies, such as Non-Volatile Memories, for increasing data locality, as in the brain.