I wrote a new blog post about Tesla and imitation learning.
The core idea:
Another idea is to train neural networks to drive via imitation learning and then run them passively in the car whenever a human is driving. Anytime the neural networks’ output is an action different from what the human driver actually did, trigger an upload. Elon Musk has alluded to Tesla’s ability to passively run self-driving software in cars, calling it “shadow mode”. The stated purpose of shadow mode is to compare the software’s output to human action. So, choosing what data to collect for imitation learning seems like a perfect application for shadow mode.
A closely related idea:
Tesla can — in theory — passively test how often the imitation networks and human drivers disagree. Once disagreements are below a certain safety threshold, the imitation networks can graduate to operating an Autopilot feature such as automatic right turns. Once that happens, Autopilot interventions can be used to trigger uploads.
What do you think? Is this a plausible approach to training?