Tesla FSD: The Most Practical AI For Navigating The Real World, According To Elon Musk
Musk also said on a side note that Tesla makes two thirds of electric vehicles in the USA (by the way, two years already without tax credit), which is in fact twice as many EVs as the rest of the automakers combined; and that Tesla is the world´s biggest robot company, as a matter of fact.
Tesla's approach to try to achieve "SAE Level 5" is to train a neural network using the behavior of hundreds of thousands of Tesla drivers using chiefly visible light cameras and information from components used for other purposes in the car (the coarse-grained two-dimensional maps used for navigation; the ultrasonic sensors used for parking, etc.). Tesla has made a deliberate decision to not use lidar, which Elon Musk has called "stupid, expensive and unnecessary". This makes Tesla's approach markedly different from that of other companies like Waymo and Cruise which train their neural networks using the behavior of highly trained drivers, and are additionally relying on highly detailed (centimeter-scale) three-dimensional maps and lidar in their autonomous vehicles.
According to Elon Musk, full autonomy is "really a software limitation: the hardware exists to create full autonomy, so it's really about developing advanced, narrow AI for the car to operate on." The Autopilot development focus is on "increasingly sophisticated neural nets that can operate in reasonably sized computers in the car". According to Musk, "the car will learn over time", including from other cars.
Tesla's software has been trained based on 3 billion miles driven by Tesla vehicles on public roads, as of April 2020; by December 2021 that number has been multiplied several times. Alongside tens of millions of miles on public roads, competitors have also trained their software on tens of billions of miles in computer simulations, at least as of January 2020; by 2021 those numbers have also multiplied in orders of magnitude. In terms of computing hardware, Tesla designed a self-driving computer chip that has been installed in its cars since March 2019 and also developed a neural network training supercomputer; other vehicle automation companies such as Waymo regularly use custom chipsets and neural networks as well.
Tesla Dojo (or Project Dojo), to which we referred the day before yesterday (in another article related to Tesla searching for talented AI engineers), is an artificial intelligence (AI) neural network training supercomputer announced by Musk on Tesla's AI Day on August 19, 2021. It had previously been mentioned by Musk in April 2019 and August 2020. According to Musk, Project Dojo will be operational in 2022.
The Dojo supercomputer uses Tesla D1 chips, designed and produced by Tesla. According to Tesla's senior director of Autopilot hardware, Ganesh Venkataramanan, the chip uses a "7-nanometer manufacturing process, with 362 teraflops of processing power", and "Tesla places 25 of these chips onto a single 'training tile', and 120 of these tiles come together... amounting to over an exaflop [a million teraflops] of power". Tesla claims that Dojo will be the fastest AI-training computer among competing offerings from Intel and Nvidia. As of August 2021, Nvidia said the current Tesla AI-training center used 720 nodes of eight Nvidia A100 Tensor Core GPUs (5,760 GPUs in total) for up to 1.8 exaflops of performance.
Musk also referred to the development of a Tesla robot during the WSJ interview, saying that it makes total sense to put all that AI learning and data science into robots. Musk is basically looking for AI engineers to develop the next generation of automation, including a general purpose, bi-pedal, humanoid robot capable of performing tasks that are unsafe, repetitive or boring.
"We’re seeking mechanical, electrical, controls and software engineers to help us leverage our AI expertise beyond our vehicle fleet", as per Tesla AI website.
You can actually apply for the available positions on this link, if interested.
All images courtesy of Tesla Inc.
Nico Caballero is the VP of Finance of Cogency Power, specializing in solar energy. He also holds a Diploma in Electric Cars from Delft University of Technology in the Netherlands, and enjoys doing research about Tesla and EV batteries. He can be reached at @NicoTorqueNews on Twitter. Nico covers Tesla and electric vehicle latest happenings at Torque News.