22-Oct-2021 | Market Research Store
Last week, Tesla revealed its plan to develop a customized chip that will be engaged in training its AI-powered autonomous driving systems within the data centers. The company has been consistently following vertical integration and the outcome of its efforts was presented on a virtual platform.
Ganesh Venkataramanan, who is a Senior Developer of Autopilot Hardware segment in Tesla, stated that the company’s Dojo supercomputer system is powered with a part of D1 chip, which uses a 7-nanometer technique and generates processing power of almost 362 teraflops. Tesla is now connecting 25 such chips into a single training tile and when 120 such tiles link together into a series of cabinets, a burst of power will be released.
Venkataramanan said that engineers at Tesla have been assembling their first cabinet very soon. He also mentioned that Tesla would have the world’s fastest Artificial Intelligence training computer. So far, Start-up company Graphcore, NVIDIA—one of the leading graphic cards manufacturers, and Intel have been manufacturing advanced processors that can be equipped within the computers or other devices to train AI models. Model training requires massive dataset management along with a broad amount of computer-based work. These chips can assist training models in identifying a group of items through visual analysis using superior cameras attached within Tesla vehicles. This sort of technology arrives after two years of sophisticated research being conducted in developing remarkable processors supporting Artificial Intelligence. The on-board software is dedicated to allowing Tesla-branded autonomous cars to respond quickly on road with the help of these chips.
Tesla recently offers a ten grand package for offering a novel self-sufficient car, which is an add-on to new vehicles. Moreover, this package would permit the car to shift lanes automatically, drive safely on highways, and effectively park at the appropriate spot. This package will hold the capacity to automatically guide city streets.
As per the critics, Tesla's driver-assistance aspects are inaccurate, as its software is not trained enough to offer level 5 autonomy, in which a car can drive in all situations without any human interference.
Read More :