The release of Tesla’s Autopilot Hardware 3.0 is just around the corner and the automotive firm has made a number of significant changes with this upgrade. A series of recent patents have confirmed that Tesla will utilize a new artificial intelligence chip, or “neural net accelerator”. This will replace the NVIDIA hardware that is currently used in the Autopilot 2.0 platform and should enable full self-driving capabilities in the future.
Why is Tesla working on a new AI chip? According to the patent filed with the United States Patent and Trademark Office, “Processing for machine learning and artificial intelligence typically requires performing mathematical operations on large sets of data and often involves solving multiple convolution layers and pooling layers.” The patent states that traditional computer processors are able to quickly perform single mathematical operation, but can only work on a small amount of data. GPU’s are able to execute multiple tasks with a larger set of data, but they were not originally designed for machine learning and AI.
Tesla therefore needed to develop a new microprocessor that could conduct “machine learning and artificial intelligence specific processing operations”. The company has created an accelerated mathematical engine, a Computational Array Microprocessor system with variable latency memory access, a Computational array microprocessor system using non-consecutive data formatting, and a Vector Computational Unit. All four systems are intended to process greater amounts of information as quickly as possible.
Tesla CEO Elon Musk recently asked employees to help test Autopilot Hardware 3.0. Employees who own a Tesla and serve as a beta tester will have the Full Self-Driving (FSD) package installed into their vehicle at no cost. Musk also remarked that Autopilot Hardware 3.0 should be available sometime within the next few months.
Autopilot Hardware 3.0’s capabilities should hopefully help reduce some of the accidents and incidents that have involved Tesla’s. In December, a California Tesla Model S driver was intoxicated and fell asleep at the wheel. Although the car was driving in a straight line, a terrible accident was still a possibility. It also took quite some time for the officers to pull the vehicle over to the side of the road.
Another driver was killed this past spring when their Tesla Model X hit a highway concrete barrier. Autopilot was engaged during the crash and the driver had not touched the wheel for at least six seconds before the accident. Tesla hopes that Autopilot Hardware 3.0's full self-driving capabilities will eliminate the potentially deadly consequences of human error.