YTread Logo
YTread Logo

Elon Musk UNLEASHES AI Powerhouse: The Dojo Supercomputer will SHOCK the Industry

Jul 07, 2023
During the iPhone launch event in 2007, Steve Jobs quoted Alan Kaye as saying that people who are really serious about software should make their own hardware and this has certainly served Apple very well, but more than a decade later , Tesla recognized that if they wanted to get ahead in software and artificial intelligence, they would need to stop depending on third-party processor chips and build their own to have control over their optimization to solve their own specific tasks and be able to scale them much faster and reach higher levels. high. of performance at a faster pace When Elon Musk has an idea like this, he prefers to tackle it on a large scale, ensuring that it is worth it for Tesla to dedicate its limited resources to a project that generates long-term benefits for the company and therefore , Elon has done exactly this by involving the teams not only with a Tesla but also with the input of Spacex to build a high-powered AI training computer called Dojo that aims to allow Tesla to achieve LeapFrog advances in the field of AI focusing on Reality.
elon musk unleashes ai powerhouse the dojo supercomputer will shock the industry
World Vision and navigation that


be used by the company's current and future products and now, two years after it was first announced, the hardware chip is finally ready for prime time and before we continue, let's stop After using Yahoo Finance, stop using Google Finance and take a look at our website, the marketplace, where we have instant stock quotes and a new set of quarterly financial data going back 15 years on more than 9 000 shares and everything is available for free at the recent Viva Tech Conference in Paris, Elon Musk reiterated to the audience that he believes that self-driving cars


soon be solved and that Tesla's value is mainly based on this basis of autonomy at first glance, This is a strange way to say it since Tesla has recently risen from almost $100. per share to the $250 level and most of the gain came over the last month, so Elon's comments seem aimed at trying to explain the recent rise by saying that Tesla stock has been boosted over the basis of autonomy.
elon musk unleashes ai powerhouse the dojo supercomputer will shock the industry

More Interesting Facts About,

elon musk unleashes ai powerhouse the dojo supercomputer will shock the industry...

Now this is possible, but arguably nothing has really changed in the last month regarding analyst opinion on fully autonomous driving, it is possible that because Tesla has been adding competitors to its Supercharger network and Elon Musk has hinted that Tesla could license other products as well, including full self-driving, there is a greater chance of this happening now; However, there are still no FSD licensing deals that we know of, especially with Ford and General Motors, which have added to the Supercharger network but have their own autonomous driving projects in the In addition to that, most analysts still They are quite skeptical about Tesla's full self-driving software as it needs to be significantly better than a human before an automated taxi business makes sense and they haven't been able to imagine Tesla's rate of improvement, plus Tesla's valuation It's even close.
elon musk unleashes ai powerhouse the dojo supercomputer will shock the industry
The $250 range where the company's market cap is around $830 billion can be justified by EV sales alone, according to Gary Black. Tesla had a much higher valuation in the past and one of the main reasons the stock has taken a hit over the past year is due to declining margins as Elon Musk aggressively reduces vehicle prices during this rising interest rate environment, investors always look ahead 6 to 12 months and therefore, as signs of reduced inflation begin to appear, the trend is likely to reverse at least slightly, which generates investors' worst fears. In Bay, Elon Musk has even said that it is more expensive to buy a Tesla with higher rates and he is counteracting this by lowering prices.
elon musk unleashes ai powerhouse the dojo supercomputer will shock the industry
This means that lower rates would mean higher prices and Tesla's margins could improve to previous higher levels, not to mention all the ongoing work Tesla has done to eliminate vehicle costs, so the market may be anticipating that the worst is over by driving the stock much higher on the basis of EV sales and margins as Tesla targets a 50% growth rate in vehicles. which currently accounts for almost all of the business's revenue along with energy, which is growing even faster than cars, then Tesla can easily grow to its now higher price-to-earnings ratio based on its current product line, in fact , autonomous Robo taxis. do not appear to be factored into the company's valuation at all, especially as forward-thinking investment firms like Arc Invest anticipate that a successful Robo Taxi launch could generate hundreds of billions of dollars in high-margin revenue in the future, what Tesla's market capitalization does not appear to reflect that the only part of the autonomy built into Tesla's market capitalization arises from its current FSD sales that are included in the price of the vehicle, but this is not related to Tesla Network Robo taxi revenue, such as charging per trip like uber.
What it does today, it's pretty obvious that if Robo taxis come to fruition and take off, Tesla's market capitalization should increase dramatically, especially if it is the only player with such a massive fleet that it can now collect the share of payment for every taxi ride Theft of a potential fleet. There are millions of Teslas already on the roads today, so Elon Musk's statement may be linking the future of Tesla's Robo taxi business to Tesla's future market capitalization, which is how he sees the path for Tesla reach a market capitalization higher than that of Apple in Saudi Aramco combined somewhat.
What he said in the past now Elon Musk has called for fully autonomous driving to be resolved in the near future several times before, but it hasn't worked, it always takes much longer than expected and has to beat local highs where it seems that the progress is being made only to level off prematurely and the desired goal is never fully achieved; However, there is one critical piece of information that suggests this time may be different. Tesla's Twitter account for Tesla AI has announced that the company is laying the groundwork not only for robotaxis. but also for robots Tesla's occupancy network, for example, which maps the 3D layout of the occupied space around its environment is transferable between cars and robots for navigation, whether on wheels or on legs.
Tesla also showed off his work on creating virtual worlds and various predictions of possible outcomes. One of the interesting things that Elon Musk highlighted was the ability to give the neural network a warning and have it perform a different action, such as changing lanes, this seems to be in line with Tesla's goal of having neural networks control the end-to-end flow by collecting multi-modal data from multiple sensors from planning to controlling the vehicle itself, being able to tell the neural network something like turn left, is a perfect use case for a human robotaxi passenger give as you go. instructions or commands to the car and make it act appropriately, so it seems that Tesla continues to close the feature gap in making Robo Taxi a commercially viable product, but Tesla has one important trick up its sleeve that could significantly advance FSD, reducing interventions and surprising. investors looking in the rear view mirror every human intervention while using fully autonomous driving is a case where the vehicle would have done something wrong on its own critical interventions are the most important ones where the human driver must intervene to avoid a collision or some other critical event, this is different from a non-critical intervention, such as taking a wrong turn or doing something that is still safe, but simply different from what the human would have wanted.
Robotaxis will only make sense if the number of critical interventions is virtually zero, while reducing non-critical ones will still improve the user experience, but is not really a requirement to see Robotaxis in action, although recent iterations of the FSD software have improved on critical issues, it could still take time to reach 99.9999 reliability. Let's say a Tesla has one of these critical events one in a hundred times, usually measured in miles or millions of miles per accident, but to keep things simple, let's say that out of every hundred events the vehicle's software needs make a decision, one of those events is a critical event in which the human needs to intervene, of course, in a robbery taxi there is no human, so this would give a reliability of 99, which is actually quite poor, especially if your car was in an accident, say one in every 100 intersections you crossed, and then reducing this to one in a thousand or one in ten thousand events is equivalent to adding nine additional 99.9 percent reliability, 99.99 percent. cent and so on, humans have about six nines of reliability 99.9998 percent now Tesla has not said. show us exactly where they are today or actually their progress in making improvements, while FSD beta testers have apparently started to see more increases without interventions, there is still a long way to go if you measure the entire fleet and so Therefore, one of the methods to improve reliability is obviously to make neural networks much stronger, so that they know how to act in a wider variety of scenarios and to avoid mistakes they may make in seemingly obvious or complex situations. , research has shown that neural networks seem to get better, more accurate, and more powerful.
They are larger and by giving them cleaner data to train on, the real world, of course, is full of infinite possibilities; However, neural networks can become surprisingly good at detecting patterns and can make decisions in fractions of a second. They have better reaction time than humans and can monitor a 360-degree view of the car's surroundings at once multiple times per second, which is why Tesla's fleet is growing exponentially as they deliver more and more vehicles each year. Even if these customers don't pay for fully autonomous driving, Tesla can still collect data from these cars, including interventions and differences between the software running in the background and what the human driver is doing, which is quite valuable for learning, as highlighted by Daniel McGuire of Arc invest.
Tesla has made some FSD software updates that have pushed FSD mileage on a rapidly exponentially growing curve relative to Waymo and Cruise, they are basically flat compared to Tesla and currently have no way to scale to a massive fleet, since there is no massive fleet of cars with everything you need. sensors in addition to Teslas Elon Musk has said that this graph will look like a wall, especially because, combined with more cars on the road, the FSD is improving, which means more people want to use it and those people want to use it for periods of time longer, so there is no shortage. of data that Tesla can collect from all over the world and they continue to train the network in all these strange scenarios making the car smarter and smarter.
The real windfall comes from Tesla's crazy Dojo training computing tutor, which allows this huge influx of data to be processed much faster than in the past, this will allow Tesla to make adjustments, test them and iterate faster on larger amounts of data. data. As of mid-last year, Tesla had the world's seventh-largest computer containing more than 7,000 Nvidia GPUs. in one of its three data centers now this is a huge computer and has brought Tesla to where it is today, however, Elon Musk has confirmed that Tesla's Dojo project will now go into production. Elon has said that it has been online and working usefully. tasks for a few months, but it looks like Tesla may be ready to start scaling the


and building exopods, which are server racks filled with Dojo chips for prime-time use, early next year.
Tesla aiming for Dojo to surpass the top 5


s in the world was combined with the assumption that they would also advance and by October 2024, which is just over a year away, Tesla's Dojo will be a staggering 40 times more powerful compared to what which is today, equivalent to over 300,000 Nvidia gpus for this game. The shifting movement that has been in the works for years makes Tesla's past progress from when they started until now look like a flat line. This could mean that a task that normally takes two days to complete will only take over an hour and therefore the quality of the processcomplete.
Autonomous driving will improve exponentially, which is what is needed, but in a short period of time, quickly add another nine or two of reliability, which is 10 to 100 times better with just the hardware, to get that kind of increase in a year is quite unprecedented. At this scale some people may look at FSD and say it's already pretty good today, however all of these exponential curves will in a sense converge next year, the number of vehicles on the road with FSD will increase and miles driven will increase. They will look like a vertical Tesla wall. is making rapid advances that stood out in the field of AI to solve problems more accurately and efficiently and the amount of computing that Tesla is adding will be an additional 100 flops per second, approximately 40 times faster than the current data center and Seemingly above the 10x gain that Elon Musk talked about during AI's second day, this may also coincide with the launch of a dedicated Robo taxi vehicle that Elon has been anticipating in 2024 during an interview with CNBC's David Faber.
Elon Musk actually says that Tesla will have its GPT chat moment this year or next year in 2024, alluding to the rapid adoption of the technology, perhaps in a matter of weeks, which is possible since Tesla's fleet already is hitting the roads today, just waiting for the software to be ready, this could instantly crush companies like Uber and Lyft and also convert third parties. match automated driving projects immediately Outdated and useless Interestingly, if Tesla is delayed in this time period, they will have more cars on the roads, which technically makes the network more powerful during this year and next year they aim for more than double the size of its entire existing fleet in Dojo terms, the exact rate of growth is difficult to predict, but the trend calls for significantly more computing to be added.
Along with further improvement of neural network algorithms, the rate of The growth that Tesla has described for Dojo will be the key to unlocking full self-driving much sooner than people think, since Elon Musk may be right this time, do you think Tesla can actually achieve it? the computing power of the


that they are looking for and do you really see full autonomous driving as a possibility by the end of next year? Don't forget to check out my latest video on Cyber ​​Truck dethroning the Ford F-150. Press the Like button. and subscribe, we would really appreciate it and a big shout out to all our sponsors who help support our channel.
Your support helps us continue creating great content. Thank you all so much for watching.

If you have any copyright issue, please Contact