Recruitment, robots, and supercomputers that are powerful
It’s has been nearly two years since Tesla’s first “Autonomy Day” event. CEO Elon Musk made numerous lofty predictions of autonomous vehicles, including his infamous claim that the company would have “one million robotaxis on the road” by the end of 2020. And now it’s time for Part Deux.
This time, the event will be called “AI Day,” and according to Musk, the “sole goal” is to persuade experts in the field of robotics and artificial intelligence to come work at Tesla. The company is known for its high turnover rate, the latest being Jerome Guillen, a key executive who worked at Tesla for ten years before recently stepping down. It has been a challenge to attract and retain top-tier talent.
The AUGUST 19TH EVENT IS SET TO START AT 5 PM PST / 8 PM ET
The August 19th event will begin at 5 PM PT / 8:00 ET at Tesla headquarters in Palo Alto (California). According to an invitation obtained from Electrek, it will include “a keynote speaker by Elon,” hardware and software demonstrations from Tesla engineers, and test rides in Model S Plaid.
Musk and other top executives at Tesla are expected to update us on Tesla’s rollout of “Full Self-Driving,” beta version 9. This beta version has been reaching more customers since this summer. We might also be able to learn about Tesla’s “Dojo” supercomputer and how it trained its neural network. Also, we may get information on the production of FSD computers. According to the invitation, there will also be an inside look at “what’s next for AI in Tesla beyond our vehicle fleet.”
Let’s start from what we know, then let’s speculate about what’s ahead.
The big news out of Tesla’s first Autonomy Day was introducing the company’s first computer chip, a 260 square millimeter piece of silicon that Musk described as “the best chip in the world.” Initially, Musk had claimed that Tesla’s cars wouldn’t need any hardware updates, only software, on the road to full autonomy. It turns out that wasn’t exactly the case; they would need this new chip — two of them, actually — to eventually drive themselves.
There has been a lot of change between the 2019 event in April and now. Last month, Tesla began shipping over-the-air software updates for FSD beta v9, its long-awaited, definitely not autonomous, but indeed advanced driver-assist system. Tesla owners who have bought the FSD option (now at $10,000) will finally be able to use many advanced Autopilot driver-assist features, such as Navigate on Autopilot and Auto Lane Change. Summon, and Traffic Light and Stop Control are also available.
TESLA’S VEHICLES ARE NOT FULLY AUTONOMOUS AFTER THE UPDATE.
Tesla’s cars will not become fully autonomous by the update. Musk also predicted that it would launch “a million” self-driving vehicles on the roads. Tesla owners who are fully self-driving need to be attentive to the road and keep their hands off the steering wheel. Some don’t, which can have tragic consequences.
FSD software has been a controversial topic recently. It is beloved by safety advocates but loathed and loved by fans. In recently publicized emails between Tesla and California’s Department of Motor Vehicles, the company’s director of Autopilot software made it clear that Musk’s comments (including his tweets) do not reflect the reality of what Tesla’s vehicles can do. And now Autopilot is under investigation by federal regulators who want to know why Teslas with Autopilot keep crashing into emergency vehicles.
FSD beta version 9 was not the only thing Tesla had to address the global chip shortage. In a recent earnings call, Musk said that its engineers had to rewrite some of their software to accommodate alternate computer chips. He stated that Tesla’s future growth would depend on a quick resolution to the global shortage of semiconductors.
Tesla depends on chips to power everything, including its airbags and the modules that control the vehicle’s seatbelts. It is not clear if the shortage has affected the FSD chips, which Samsung produces. Musk and his colleagues may be able to provide some insight during this week’s event.
Tesla uses a supercomputer to train its AI software, fed to its customers through over-the-air software upgrades. Musk called this super-powerful training computer “Dojo” in 2019, referring to it as.
“Tesla is developing a [neural net] training computer called Dojo to process truly vast amounts of video data,” he later tweeted. “It’s amazing!”
He also mentioned Dojo’s computing capabilities, saying it could perform an exaFLOP (1018 floating-point operations per sec). This is a fantastic amount of power. NetworkWorld stated last year that “to match what a one-exaFLOP computer can do in one second,” they wrote, “you would have to perform one calculation per minute for 31,688,765,000 Years.”
“IT’S A BEAST!”
Cray, a chipmaker AMD, and Cray, a computer builder, are currently working with the US Department of Energy to create the world’s most powerful supercomputer. It will have 1.5 exaFLOPs of processing speed. Dubbed Frontier, AMD says the supercomputer will have as much processing power as the next 160 fastest supercomputers combined.
Dojo, when completed, is expected to be one of the most powerful supercomputers in the world. Tesla’s supercomputer runs a neural net to train its AI software to drive self-driving cars, rather than doing complex calculations in areas like climate and nuclear research. Musk said that Dojo would be available to all companies who want it to train their neural network.
Andrej Karpathy was Tesla’s chief of AI earlier this year. He gave a presentation to the 2021 Conference on Computer Vision and Pattern Recognition. During his presentation, he provided additional details about Dojo and its neural network.
According to Electrek Karpathy stated, “Computer vision is for us the bread and butter in what we do” and that it enables Autopilot. “And that’s why we have to master the data from our fleet and train huge neural nets. We also need to experiment a lot. The compute was a significant investment.
Earlier this month, Dennis Hong, founder of the Robotics and Mechanisms Laboratory at UCLA, tweeted a photo of a computer chip that many speculate is the in-house hardware used by Tesla’s Dojo.
Hong is attractive for many reasons. He specialized in humanoid robots and participated in the DARPA Urban Challenge, which kicked off the race for self-driving cars. His team came in third.
Hong responded to Twitter’s question on whether his lab was working alongside Tesla. We will likely learn more about Hong’s and Tesla’s intersections during AI Day.
Musk has made it clear that he wants Tesla to be more than just a car manufacturer. He stated that people would view Tesla as an AI robots company over the long term.
For anyone watching the AI Day Livestream, a warning: Musk’s predictions on near-term accomplishments should be taken with a large grain of salt. It is unlikely that the topics discussed at this event will have any significant impact on its business in the coming months.
The task of creating self-driving cars is challenging. Even companies like Waymo that are perceived to have the best autonomous vehicle technology are still struggling to get it right. Tesla is no exception.
“A key question for investors will be what the latest timeline is for achieving full autonomy,” Loup Funds managing partner Gene Munster said in a note. Despite Elon’s ambitious goal to finish this year, we believe that 2025 will be the first year public access to level 4 autonomy.
Tesla is already busy in 2021. The company will need to open new factories in Texas and Germany. And it needs to tool up production for its hotly anticipated Cybertruck, which has been delayed until 2022. Full autonomy can wait, as it is.