Some of you might be interested in paper I found titled "End to End Learning for Self-Driving Cars." Its authors all work for Nvidia. Some background:
The Tesla Model S and X have a Nvidia Tegra processor, which includes a very powerful GPU, Graphical Processing Unit. GPUs can efficiently perform complex matrix operations that may have little to do with graphics many times faster than a conventional processor. Nvidia has built their company for the gaming community, but it now supplies software for their GPUs that implement Deep Learning Neural Networks, which they claim shows strong potential for Autonomous Driving.
Several years ago Neural Networks were resurrected from a long period of decline with the discovery of new methods for training deep (many layered) neural networks.
The paper describes the architecture of a Deep Learning Neural Network as implemented by Nvidia, and it gives a hint to the processing involved in a neural network powerful enough to drive a car autonomously.
Tesla has never said whether their "assisted driving" system uses Deep Learning Neural Nets. To my knowledge they have not even said to what extent they use the Tegra processor for things other than just the graphics on the Tesla 17" display. They have said that they do not just use the standard Mobileye processing that comes with the Mobileye camera. Since the Tesla has a radar as well as the camera, It is doing some sensor fusion beyond what the Mobileye camera can do alone.