Almost two years have passed since Tesla’s first “Autonomy Day” event. At the event, CEO Elon Musk made numerous sublime predictions about the future of autonomous vehicles, including his infamous claim that the company will have “a million robotaxi on the road” by the end of the year. And now it’s time for Part Deux.
This event will be held as ‘AI Day’, According to Musk, “the only goal” is to persuade experts in robotics and artificial intelligence to work for Tesla. The company is known for its high turnover rate, most recently with Jerome Guillen, a key executive who served at Tesla for 10 years before leaving the company. Attracting and retaining talent, especially first-class names, has proven to be a challenge for the company.
It will kick off on August 19 at 5pm PT / 8pm EST at Tesla’s headquarters in Palo Alto, CA. According to an invitation from electric, “Keynote by Elon, hardware and software demonstrations by Tesla engineers, test drives of the Model S Plaid, and more.” Like Battery Day, this event will be broadcast live on the Tesla website, giving investors, media and many fans of the company a closer look at what’s in development.
Musk and other senior officials at the company are expected to provide updates on Tesla’s “Fully Autonomous Driving” (FSD) beta 9 release, which is starting to reach more customers this summer. You can also get details about Tesla’s “Dojo” supercomputer, training neural networks, and production of FSD computer chips. And there will also be “an inside look at the next level of AI at Tesla, beyond our vehicles.”
Let’s start with what we know and move on to guesses about what’s going to happen.
The big news for Tesla’s first Autonomy Day was the introduction of Tesla’s first computer chip, a 260 square millimeter piece of silicon that Musk described as “the best chip in the world.” Originally, Musk argued that Tesla’s cars did not require hardware updates for full autonomy, only software. That wasn’t exactly true. They will eventually need these new chips (actually two of them) to drive themselves.
A lot has happened between the 2019 event and now. Last month, Tesla started delivering the long-awaited over-the-air software update for FSD Beta v9. That said, Tesla owners who have bought the FSD option (currently priced at $10,000) will finally have Navigation on Autopilot, Auto Lane Change, AutoPark, Summon, Traffic Light and Stop Control.
This update won’t make Tesla’s cars fully autonomous, nor will it launch “a million self-driving cars” on the road as Musk predicted. Tesla owners with fully autonomous driving capabilities still need to pay attention to the road and hold the steering wheel. Some don’t, which can have tragic consequences.
Loved by fans and hated by safety advocates, FSD software recently plunged Tesla into hot water. In a recently published email between Tesla and the California Department of Motor Vehicles, the company’s Director of Autopilot software made it clear that Musk’s comments (including tweets) do not reflect the reality of what Tesla vehicles can actually do. And now Autopilot is under investigation by federal regulators who want to know why Tesla with Autopilot keeps crashing into emergency vehicles.
In addition to the FSD beta v9 launch, Tesla had to adapt to a global chip shortage. In a recent earnings call, Musk said the company’s engineers would have to rewrite some software to accommodate replacement computer chips. He also said that Tesla’s future growth will depend on a swift resolution to the global semiconductor shortage.
Tesla relies on chips to power everything from airbags to the modules that control the vehicle’s seat belts. It’s unclear whether Samsung’s FSD chips are being affected by the shortage. Musk and his colleagues can provide insight into this at this week’s event.
Tesla uses powerful supercomputers outside the vehicle to train its AI software and then deliver it to customers via over-the-air software updates. In 2019, Musk teased this “super-powerful training computer” and called it “Dojo.”
“Tesla [neural net] “It’s training a computer called Dojo to process really massive amounts of video data.” he later tweeted. “Beast!”
He also hinted at Dojo’s computing power, claiming that Dojo is capable of exaFLOP or 1 quintillion per second (1018) floating-point operations. Unbelievable power. “To match what one exaFLOP computer system can do in just one second” network world I wrote last year, “You have to do a calculation once per second for 31,688,765,000 years.”
By comparison, chip maker AMD and computer maker Cray are now working with the US Department of Energy to design the world’s fastest supercomputer with 1.5 exaflops of processing power. Called Frontier, AMD says its supercomputers will have the same processing power as the next 160 fastest supercomputers put together.
When completed, Dojo is expected to be one of the most powerful supercomputers on the planet. But instead of performing advanced computations in areas like nuclear and climate research, Tesla’s supercomputers are running neural networks to train AI software to power self-driving cars. Ultimately, Musk said Tesla will offer Dojo to other companies that want to use it to train neural networks.
Earlier this year, Andrej Karpathy, head of AI at Tesla, gave more details about Dojo and neural networks at the 2021 Computer Vision and Pattern Recognition conference.
“For us, computer vision is the bread and butter that makes Autopilot possible and what we do,” Karpathy said. electric. “And for it to work really well, you have to master the fleet’s data, train massive neural networks and experiment a lot. So we’ve invested a lot in computing.”
Earlier this month, Dennis Hong, founder of the UCLA Institute of Robotics and Mechanics, said: Tweeted a picture of a computer chip Many speculate that it is the in-house hardware used by Tesla’s Dojo.
But Hong is also an interesting person for other reasons. He specializes in humanoid robots and took part in the DARPA Urban Challenge, which launched a self-driving car competition. (His team finished third.)
Asked on Twitter if his lab was working with Tesla, Hong posted a playful emoji, but if not, declined to comment. During AI Day, you can learn more about how Hong’s work intersects with Tesla’s pursuit.
Musk opened up about his aspirations for Tesla to be more than just a car company. “In the long run, people will think of Tesla as an AI robotics company as much as an auto or energy company,” he said earlier this year.
A warning to everyone participating in the AI Day livestream: Follow Musk’s predictions of near-future achievements with huge amounts of salt. What will be discussed at this event will not have a measurable impact on the company’s business for months to come.
Self-driving cars are a real challenge. Even companies like Waymo, which are perceived to have the best autonomous vehicle technology, are still struggling to get it right. Tesla is no different.
“The most important question for investors is when will the latest timeline to achieve full autonomy be,” said Gene Munster, managing partner at Loop Fund, in a memo. “Despite Elon’s ambitious goals for the end of the year, our best guess is that 2025 will be the first year that Level 4 autonomy will be publicly available.”
The rest of 2021 is already full for Tesla. The company should open factories in Texas and Germany. And to live up to Cybertruck’s hot expectations, postponed to 2022, production must be instrumented. Full autonomy, as it is now, can wait.