Partnering with Purdue on Project AXON

Underneath electrochemical processes and dendrites, optical spin channels are one area that cells communicate. Understanding both layers can be used to develop better chips

I am pleased to announce my partnership with Purdue on Project AXON. What I hope to glean from this partnership in coordination with Project BLUEBIRD is to develop newer chips that effectively mimic the way in which the brain processes information so efficiently. While Project BLUEBIRD explores optical/spin communication channels between cells, Project AXON explores the abstraction layer right above that.

Most low power memory is volatile, so cutting the power erases what the system learned and it starts over from nothing. AXON proposes magnetic memory instead, which holds its state with the power off and shrugs off radiation doses far beyond anything a mission would deliver. On a satellite that cycles its power, that is the difference between a system that resumes and one that restarts.

Neuromorphic chips are built loosely on the way biological neurons work. Instead of separating memory and processing the way an ordinary computer does, they store information in the same place they compute with it, which cuts out a great deal of the energy normally spent shuttling data back and forth.

Concept flow chart.

There is a second reason this matters in space. Ordinary working memory forgets everything the moment power is interrupted, so a payload that reboots has to reload what it learned across a slow connection. Some of the device technologies we are evaluating store information magnetically, which means the state survives a power cycle.