Tesla AI Day 2022: Optimus robot, Full Self-Driving, supercomputers, and more from Elon Musk’s big event

Tesla’s AI Day event kicks off September 30th, during which the company is expected to highlight its ongoing efforts to develop its advanced driver-assist features that Elon Musk has promised will eventually lead to fully autonomous cars. 

But this year, Tesla’s cars and self-driving projects are likely to take a backseat to a robot named Optimus. Musk announced the “Tesla Bot” at last year’s AI Day, promising that it would be “friendly” and potentially revolutionize the company’s assembly line and manufacturing business. 

Since then, Musk has gone on to hype the robot as “the most important product development we’re doing this year,” predicting that it will have “the potential to be more significant than the vehicle business over time.” Future applications could include cooking, gardening, or even “catgirl” sex partners, Musk said. Production could start as soon as next year. 

To say nothing of the fact that Musk announced the robot by bringing out a person clad in a spandex costume to dance awkwardly onstage, there’s a lot of room for skepticism. Whether the robot is real or just a ruse to gin up some excitement about Tesla’s AI division that has seen a lot of turnover in recent years probably won’t be decided after Friday’s event. But be warned: what you see is not necessarily what you get. 

With that said, let’s dive into what we should expect from Tesla’s upcoming AI Day.  

To Musk, the idea of building a humanoid robot is a no-brainer. 

“Tesla is arguably the world’s biggest robotics company because our cars are semi-sentient robots on wheels,” he said during last year’s AI Day. “It kind of makes sense to put that onto a humanoid form.”

To be sure, Musk’s cars are not “semi-sentient”; they require constant monitoring by a human driver in order to operate safely. Even then, they tend to crash into stationary objects with enough frequency to spur multiple investigations from federal agencies. 

Nonetheless, Musk said the Tesla Bot would be “friendly,” with “human-level hands,” “Autopilot cameras” for eyes, and a “Full Self-Driving computer” for a brain. The robot would measure 5’8’’ and weigh 125 pounds, with the ability to carry up to 45 pounds in its “human-level hands.”

The Tesla Bot has the potential to “be more significant” than Tesla’s car business, Musk said. It will be in people’s homes, though it will be more than 10 years before it’s ready to be sold to consumers. People will buy them for their aging parents as a gift, he predicted. 

“This will be quite profound,” Musk added.

In June, Musk tweeted that AI Day would be pushed to the end of September (it was originally scheduled for August) to allow more time to show off a working prototype. Later, he tweeted an image of robotic hands forming a heart shape to promote the upcoming event. 

But let’s put aside Musk’s bluster for now and look at the actual evidence of Tesla’s robotics program. For over a year, Dennis Hong, founder of the Robotics and Mechanisms Laboratory at UCLA, has been teasing a collaboration with Tesla, but we’ve yet to see any confirmed reports of that. For now, it seems as though Hong is content to tweet mysteriously about Tesla’s robotics program and leave it at that. 

What is happening is Tesla is going on a hiring spree for motor and actuator designers and engineers to work in its robotics department. Now that Musk has declared Optimus a top priority, his company is scrambling to hire enough staff to meet that directive. Musk has even directed the company’s Autopilot team to shift their focus to Optimus to meet the current deadline. 

“Developing a HUMANOID ROBOT requires pushing the technology boundaries even further, bringing us new challenges,” Konstantinos Laskaris, Tesla’s longtime lead electric motor designer, said in a LinkedIn post announcing new open positions. 

Experts note that no company on earth is close to building a robot with the same mobility and dexterity as a human. Even Boston Dynamics, which makes one of the most advanced bipedal robots in the world, says its Atlas robot is just a research platform at this point — not a consumer product. Atlas still frequently trips and falls, despite over a decade of work.

Parsing jokes from reality has never been Musk’s strong suit, and the gap between what’s real and bluster has never been wider than it is with this robot. Tesla’s history is littered with fanciful ideas that never panned out, like a solar-powered Supercharger network, battery swapping, robotic snake-style chargers, city-to-city rocket travel, or a self-driving car you could summon from across the country. 

The Tesla Bot could certainly end up sucked into the memory hole along with those other pie-in-the-sky ideas. But it seems unlikely Musk would easily part with the idea of Tesla as a robotics company, especially since so much of the company’s value is directly tied to his appearance of being on the vanguard of technological innovation. 

While the Tesla Bot is sure to steal all the headlines, new details about the company’s driver assistance technology are what most people will be tuning in for. Tesla has said that more than 160,000 drivers in the US and Canada are using its Full Self-Driving (FSD) beta system, which despite its name, does not enable the vehicle to drive autonomously. (FSD is not approved for use in the EU, where Tesla must clear a higher regulatory bar to get the green light.)

Musk has been promising actual driverless cars are coming since 2016 — a promise he has yet to deliver. He’s gone from saying that Tesla will have 1 million robotaxis on the road by the end of the year to 1 million people in the FSD beta program, which are wildly different things.

Loved by fans and loathed by safety advocates, the FSD software has gotten Tesla in a lot of hot water recently. The National Highway Traffic Safety Administration is currently investigating 16 crashes in which Tesla vehicle owners using Autopilot crashed into stationary emergency vehicles, resulting in 15 injuries and one fatality. Most of these incidents took place after dark, with the software ignoring scene control measures including warning lights, flares, cones, and an illuminated arrow board. The probe was recently upgraded to an “Engineering Analysis,” which is the second and final phase of an investigation before a possible recall.

Tesla vehicles today come standard with a driver-assist feature called Autopilot. For an additional $15,000, owners can buy the FSD option, which Musk has repeatedly promised will one day deliver fully autonomous capabilities. But to date, FSD remains a “Level 2” advanced driver-assistance system, meaning the driver must stay fully engaged in the operation of the vehicle while it’s in motion.

The company has been accused of false advertising by regulators and sued by customers for allegedly misleading them about the capabilities of their vehicles. But FSD is also crucial to Musk’s vision to portray Tesla as a leader in AI and robotics. And Musk has largely avoided any serious consequences in his pursuit to be at the bleeding edge of technology.

Outside of the car and the robot, AI Day will be an opportunity for Tesla to update investors and fans on its computational advances. Tesla uses a powerful supercomputer to train the company’s AI software that then gets fed to its customers via over-the-air software updates. More specifically, it processes all the video feeds from Tesla’s fleet of over 1 million camera-equipped vehicles on the road today. 

The automaker already has a large Nvidia GPU-based supercomputer that is one of the most powerful in the world, but the new Dojo custom-built computer is using chips designed by Tesla. In 2019, Musk gave this “super powerful training computer” a name: Dojo.

Musk has claimed that Dojo will be capable of an exaFLOP, or one quintillion (​​1018) floating-point operations per second. That is an incredible amount of power. “To match what a one exaFLOP computer system can do in just one second,” Network World wrote, “you’d have to perform one calculation every second for 31,688,765,000 years.”

Tesla is not alone in developing powerful computers. Chipmaker AMD and computer builder Cray are currently working with the US Department of Energy on the design of the world’s fastest supercomputer, with 1.5 exaFLOPs of processing power. Dubbed Frontier, AMD says the supercomputer will have as much processing power as the next 160 fastest supercomputers combined.

At last year’s AI Day, Dojo was still a work in progress. Executives revealed its first chip and training tiles, which would eventually develop into a full Dojo cluster or “ExaPod.” Tesla said it will combine 2 x 3 tiles in a tray and two trays in a computer cabinet for over 100 PFlops per cabinet. In a 10-cabinet system, Tesla’s Dojo ExaPod will break the barrier of the ExaFlop of compute. 

When completed, Dojo is expected to be among the most powerful supercomputers on the planet. But rather than performing advanced calculations in areas like nuclear and climate research, Tesla’s supercomputer is running a neural net for the purposes of training its AI software to power self-driving cars. Ultimately, Musk has said Tesla will make Dojo available to other companies that want to use it to train their neural networks.

If last year is any indication, the Dojo presentation is likely to get very technical, but the hope is that it will help Tesla recruit more engineers to work on its computing program. The company has seen a lot of turnover recently, including Andrej Karpathy, who led Tesla’s Autopilot team and AI division. His departure came soon after the closure of a Tesla office in San Mateo, California, where data annotation teams were helping to improve the company’s driver assistance technology. According to public records, over 220 people were dismissed from their positions. 

A warning for anyone tuning in to the AI Day livestream: take Musk’s predictions about near-term accomplishments with a massive grain of salt. The things that will be discussed during this event are unlikely to have any measurable impact on the company’s business in the months to come. 

Self-driving cars are an incredibly difficult challenge. Even companies like Alphabet’s Waymo that are perceived to have the best autonomous vehicle technology are still struggling to get it right. Tesla is no different.

The Tesla Bot is sure to generate a lot of interest, but it also should not be mistaken for anything close to production-ready. As The Verge’s James Vincent aptly noted last year, Musk has a reputation for the “bait-and-switch.” Just look at the hyperloop: what was first described as an ultrafast system perfect for mass transit has instead turned into a small tunnel designed just for cars. 

Tesla’s AI Day is a good opportunity for the company to draw attention away from the litany of headaches that it’s currently dealing with: fires at its Megapack facility in California and its Berlin Gigafactory; multiple overlapping investigations into its driver assistance technology; shrinking market share as more legacy automakers shift to EV sales; and Musk’s attempts to wriggle out of buying Twitter.

AI Day will be a chance for Tesla to focus on the future, even as its present looks increasingly precarious.