What is Tesla's supercomputer Dojo？
Abstract
Teslabegan production of its Dojo supercomputer in July. The system is designed to trainartificial intelligence(AI) models for autonomous driving cars, and Tesla plans to invest over $1 billion in Dojo by next year.Morgan Stanley analysts have raised their target price for Tesla by 60% to $400, and compared Dojo to Amazon's profit growth engineAWS.
They expect that Dojo will accelerate the development of autonomous driving technology and save costs of up to $6.5 billion in the next few years. This represents a significant change in the market's perception of Tesla.What magical abilities does Tesla's Dojo supercomputer have that can unlockAImagic, as Morgan Stanley suggests?
What is Tesla’s powerful supercomputer Dojo?
Tesla's Dojo is a powerful supercomputer designed by the electric vehicle manufacturer to accelerate the development of autonomous driving technology, among other applications. The system was first announced in Tesla's AI Day event in 2021 and is set to become one of the world's most powerful computing clusters when fully operationalTo understand Dojo, we need to start with FSD. Full self-driving is a crazy idea - but how can it be achieved? You can think of it this way: why can humans drive cars? It's thanks to our eyes, brain, and hands/feet working together.
So, if humans can drive, can't we apply that same approach using cameras, AI, and vehicle controls? Cameras act like eyes to observe the surroundings, AI makes decisions like a brain, and vehicle controls respond like hands/feet to navigate the environment seamlessly.
The question arises: what is AI, and where did it come from? It can't just be a figment of our imagination. Simply put, AI is essentially a simulation of the human brain that, through long-term observation, learning, induction, reasoning, verification, and so on, has developed a series of experiences to cope with different situations.
So how can AI learn these things?
The answer is "deep learning." Deep learning involves feeding two sets of data into a computer - one for it to learn using algorithmic self-exploration, and another which contains the correct answers. The machine compares its results to the correct answers, learns from its mistakes, and repeats the process until it has developed an accurate response pattern. This sounds like the same process humans go through to learn something, doesn't it?However, even with deep learning, there is still a problem when dealing with an increasing amount of data.
The computational power required may become overwhelmed by the speed at which data is inputted, leading to a situation where input exceeds output and calculations cannot keep up. This is where Tesla's Dojo supercomputer comes in - it was designed to provide enough computing power to handle large amounts of data efficiently and quickly.This is a powerful computing center with the ability to handle large amounts of AI tasks, such as video processing and complex model training. Its design is intended to process large amounts of video data, accelerating the iteration of Tesla's Autopilot and Full Self-Driving (FSD) systems, while also providing computational support for Tesla's humanoid robot Optimus.
What are the advantages of dojo’s computing power?
Dojo is designed to handle large amounts of data for training driving systems, making it particularly suitable for use in developing autonomous driving software. It also has applications in video processing, complex model training, and providing computational support for Tesla's humanoid robot Optimus.The custom computer useschipsdesigned by Tesla and the entire infrastructure, setting it apart from Tesla's previous supercomputer based on Nvidia GPUs.
Tesla plans to increase the scale of Dojo until the cluster becomes one of the world's most powerful supercomputer clusters by early 2024.The system consists of 120 training modules, each with 3,000 D1chips, totaling over one million training nodes, with a total computing power of 1.1EFLOP (10,000 trillion floating-point operations per second). Tesla's chief engineer, Tim Zaman, stated that their computing cluster has only 0.3% idle time, with 84% of the time spent on high-priority tasks.
What significant impact has supercomputing had on Tesla?
The potential benefits of Dojo for Tesla are significant. The supercomputer can help improve the efficiency, speed, and cost advantages of developing autonomous capabilities, with the potential to shorten the time required to develop such technologies significantly.
Better FSD products would bring higher profits and revenues to Tesla, as customers decide to add FSD software functionality when they purchase Tesla vehicles.Seth Goldstein, Morningstar Research Services' stock strategist, predicted on Monday that if Dojo is launched, operated, and takes over software training, it will allow Tesla to update software more quickly and increase software profit subscriptions, which could be a huge value driver.
Disclaimer: Community is offered by Moomoo Technologies Inc. and is for educational purposes only.
Read more
Comment
Sign in to post a comment
TeslaSmurf : Dojo is ONE of the TOOLS Tesla have, as the 10.000 NVDIA H100 just went on line for the FSD Ai training program. The point is not Dojo (a Tesla-built mega computer: one of the 5 most powerful in the world) which is specific for the vision Ai, but FSD itself.
People can’t still realise that it will change our world, not just with self driving cars, but with FSD driven Tesla robots “seeing” the real world (instead of mere data). No one else will have that in the next 10 years and when someone else gets to it, Tesla will have been kept innovating at a faster speed, reason why there’s NO POSSIBLE catching up Tesla.
ZnWC : Here is the article on how Morgan Stanley optimistic view and arrived at Tesla share target price USD 400:
Tesla (TSLA) stock surges from optimistic look at Dojo supercomputer
71518529 : Thanks for the info on DOJO .. and Thank you for keeping us Tesla investors informed..
Ava Quinn OP 71518529 : So glad to hear that~~~thx