Engineering

Neuromorphic chips mimics organic neural networks, offering much higher power efficiency compared to more classical architecture. Spiking neurons, a counterpart to logic gates, activate only when receiving sufficiently strong signal. 

Neuromorphic architecture allows for bidirectional communication between units enabling processing of independent information stream, a low-level software concept, on hardware level. On top of that, communication happens between related units rather than going through limited number of shared buses as they do in today's computers.

Below are diagrams of two architectures: Von Neumann's (classical) and Neuromorphic (theorized)

Here's low-tech article from 2017 about neuromorphic computing

Engineering @neynar. Amusement, wonder, and uncertainty ✴︎ libred.org

Here's low-tech article from 2017 about neuromorphic computing https://spectrum.ieee.org/neuromorphic-chips-are-destined-for-deep-learningor-obscurity