When you watch Tesla debut its Optimus humanoid robot at AI Day 2022, it’s entirely fair to be skeptical about the company’s big plans. The robot’s gliding gait was difficult compared to the exciting parkour and flips of Boston Dynamics’ Atlas, and Tesla missed many deadlines to fully implement its self-driving technology in its cars.
But I was at the AI Day event, and I’m here to warn you against dismissing Tesla’s sci-fi-inspired vision.
The company has revealed plenty of evidence that it’s thinking deeply about a bipedal robot and the AI technology needed to make it useful. And much of Tesla’s history of true excellence in electric vehicle engineering and manufacturing relates to the Optimus.
Non-Tesla robotics and AI experts I spoke to at the event were impressed with Optimus, the Tesla Bot’s codename. And Tesla’s engineers had the kind of fire in their belly that portends progress, especially when backed by CEO Elon Musk’s vision and Tesla’s resources.
“I haven’t worked this hard since undergrad. But I like it,” said one senior chip designer working on Tesla’s Dojo technology to train the AI systems behind Optimus’ ability to navigate rooms and Tesla’s FSD technology for driving cars. He was not authorized to speak to the media.
I was also impressed compared to last year when Tesla debuted Optimus as a few presentation slides accompanied by a dancer dressed as a robot.
I didn’t cheer and applaud like many of the 1,000-plus people packed into a giant store at Tesla’s offices in Palo Alto, California, when the first Tesla Bot walked across the stage. I really appreciated the engineering feat when I saw the computer-controlled, non-rolling bipedal waist hinge, powered by a computer on his chest, powered by a collection of shiny cylindrical actuators studded with status LEDs and wired with a copper nervous system. I heard its cooling fans blow and later strained to lift the 42-pound block of copper and electronics that makes up each 25-processor Dojo AI training tile.
It’s not yet clear whether humanoid robots will one day do the shopping or take the place of humans working on production lines, as Musk envisions. The controversial and outspoken billionaire estimates it will be three to five years before the first Optimus goes on sale, an eternity in the tech world. But if his vision is true, it could signal a transformation of the world as profound as the automobile or the smartphone.
“We wouldn’t be surprised to see it become a major driver of the share price towards the end of the decade,” New Street Research analyst Pierre Ferragu said in a report on Monday. “Imagine Optimus as a startup today: it could be valued at several billion dollars, maybe even several tens.”
There’s certainly a long journey from prototype to product, as AI Day attendees were reminded when they saw the Tesla Semi and Cybertrucktwo vehicles they have missed original shipping dates.
While Tesla’s ability to realize its robot vision is even more advanced, I’d recommend not writing the project off simply as Musk’s utopian fantasy. That’s why.
Musk gets credit for creating geek heaven
Musk has a talent for picking businesses that are difficult but achievable, finding engineers ready for the challenge, and providing them with an environment where they expect to make a difference.
One who works at SpaceX, which has upended the rocket business the same way Tesla is shaking up the auto industry, said rivals have tried to hire her several times. These would be light jobs, she said, but in practice she wouldn’t really be able to do anything there. She was not authorized to speak to the media.
We may spend a lot of money on laptops and smartphones, but Musk is looking for newer horizons. “You definitely want to see what happens with Optimus while a bunch of other technologies are kind of plateauing,” Musk said at the AI Day event.
That message resonated with an engineer I spoke to at AI Day who works on actuators, the key mechanism that moves Optimus’ torso, legs, arms and fingers. He previously worked at Boeing, but sees robotics as a new wave of innovation.
AI Day 2022 is specifically set up as an engineering recruiting event. One I spoke to who works for a direct competitor to Tesla came up because of a recruiting email. Another, who works on surgical robots, was clearly impressed by the scope and funding of Tesla’s work in the field. Nobody was laughing at Tesla’s Optimus.
Tesla has invested heavily in its Optimus robot
On stage at AI Day, a pre-Optimus prototype called Bumble-C wiggled, waved, pumped its arms and bent at the waist. It was unremarkable by some standards, but also the result of less than a year of work. A series of engineers shared the stage with Musk to detail the research done so far on Optimus. Between them:
- Tesla designs its own powertrains, tightly packed sets of gears, motors, sensors and controls that act like human muscles. The company is drawing on the team that already designs the powertrains of Tesla cars. At AI Day, engineers showed Tesla’s calculations of how it settled on a collection of six drivetrains for the entire robot that best optimize cost, ease of manufacture, speed, torque, mass and efficiency.
- Tesla took lessons from human anatomy, for example, using a complex four-bar mechanism similar to the human knee that adapts the actuator’s different needs for force or speed depending on how much the knee is bent.
- Tesla uses the same AI technology behind FSD to drive Optimus as well. This includes the Occupancy Network, an AI system that converts camera input into a 3D map of everything around a robot.
- Tesla is building its own custom data center technology, Dojo, to train multiple AI systems that go into a vehicle or robot. So far, Tesla has built three cabinets of Dojo hardware and can now process video training data in ways not possible with the earlier approach, using 72 of Nvidia’s high-end A100 processors.
- For locomotion, Optimus combines a physical model that simulates a virtual robot with data from real-world sensors that measure how the robot actually performs. The technology of walking became more complex: the first steps were in February; Optimus’ pelvis began to swing in July; a swinging motion of the arms accompanying the steps, beginning in August; and Optimus’ toes first lifted off the ground in September.
“The Tesla team is so forward-thinking and so confident in its pace of innovation that it doesn’t shy away from sharing details about what it’s doing,” Ferragu said.
Tesla’s Optimus robot requires AI
Optimus faces a challenge that Tesla’s cars don’t: it requires AI for everything it does. Tesla can make a successful business selling industry-leading fully human-driven electric vehicles, then gradually develop autonomous vehicle technology.
In September, Tesla expanded beta testing of FSD, its software designed to eventually lead to self-driving cars. 160,000 people in the US and Canada have access, CEO Elon Musk said.
Stephen Shankland/CNET
With humanoid robots, the AI should work from the start. And the AI challenge is harder with robots.
A brief example of today’s AI technology: Instead of programming computers with narrow, rigid if-this-then instructions, AI works by training a system to recognize patterns in vast amounts of real-world data. This allows AI systems to handle much more complexity and make more nuanced decisions.
The hard part is that self-driving cars encounter a huge variety of situations. Even the same road intersection can be very different if it’s raining, under construction, or blocked by a stalled car.
Tesla is investing heavily to overcome these challenges, mining 100 terabytes of video data reported by its cars every day, according to one engineer at Tesla AI Day. It also simulates different conditions to extend AI training situations beyond this real-world data.
But the variety of situations that robots can face is significantly wider. Consider how different one home is from the one next door. Then compare these homes to businesses, sidewalks, and farms.
Tesla’s robot demonstration videos, however, were set in a relatively narrow area: its offices and research labs. Musk said the company plans to test them in its own “gigafactories” first. This can provide enough training data to give robots useful support.
Why create Optimus the robot?
Musk offered a somewhat murky motivation for why Optimus was built. SpaceX’s job is to get humanity to Mars, and Tesla’s electric vehicles and batteries are supposed to wean us off fossil fuels. Musk’s explanation for Optimus seemed more opportunistic: Tesla has the experience, so it can do it.
Musk hopes Optimus will “help millions of people” by freeing people from boring, dangerous and repetitive work. At his most exciting, he said robots could lead to “a future of abundance, a future where there is no poverty, where people can have whatever they want in terms of products and services. This is truly a fundamental transformation of civilization as we know it.”
I wouldn’t expect an end to poverty any time soon. But today’s Optimus takes big steps, literally and figuratively, from last year’s debut as just an idea.