Tesla’s Neural System Networks for Self-Driving Improved

Tesla’s head of AI and computer vision, Andrej Karpathy, explains how the new neural networks for self-driving works. Through talks, he explains the system to the investors and owners to build confidence about the capability of Tesla´s self-driven system.

Tesla uses thousands of microchips known as graphics processing units, or GPUs, to train many simultaneous networks at its headquarters. In addition to receiving regular data from cameras and car radars currently in transit, members of the autopilot team may request specific information collected. This information includes when cars are diverted from bicycles or trucks, by training networks for the records they use to these situations.

Tesla has an advantage over other autonomous car developers: It has sold hundreds of thousands of cars, each connected to the internet, to use their expertise to “train” autopilot and software to know what is working and how it works. Data enters into an intelligence system that recognizes patterns, especially other vehicles, and how they move, and that speeds up over time.

The video below shows the car looking for its owner:


Leave a Reply