Tesla hired engineers to make a robot for dangerous or boring work. At AI Day, expect updates to FSD technology for possible self-driving cars as well.
Tesla, which continues to work on creating humanoid robots and building truly self-driving cars, will detail its work at the company’s second AI Day, which begins at 5 p.m. PT Friday. Expect to see some of the most advanced examples of AI technology around, but be prepared to wait a few more years before this technology is ready for the mainstream.
The likely star of the show is the Tesla Bot, a humanoid robot codenamed Optimus that CEO Elon Musk debuted at the first AI Day. Last year we only saw specs, a dummy mockup and a wild dance from someone wearing Optimus gear in a strange attempt to illustrate what the Tesla Bot would look like. Musk postponed the AI for the next day to wait for the physical prototype of the Tesla Bot.
Previous events in this vein, such as last year’s AI Day and the debut of Neuralink in 2019, have featured cutting-edge technology. But they were also aimed at helping Musk recruit ambitious and talented engineers, and day two AI follows that pattern.
“This event is for recruiting AI and robotics engineers, so it will be highly technical,” Musk tweeted Thursday. “AI/robotics engineers who understand what problems need to be solved will like what they see,” he added.
Tesla is hiring Tesla Bot engineers to build not only walking humanoid robots but also wheeled models for factories, according to job postings spotted by Reuters, and Musk is explicitly planning cool events like AI Day to recruit employees.
Musk himself has worried about how small we’ll be compared to the super-intelligent AIs that pose an “existential threat” to humanity, so don’t feel silly if you’re afraid to bend the knee to our future robot overlords. However, with Optimus, Tesla is trying to ease our minds as it sees the Tesla Bot as a useful helper. On his Instagram page, he showed a picture of robotic hands forming a heart shape and promised: “If you can run faster than 5 miles per hour, you’ll be fine.”
Sharing the spotlight with the Tesla Bot is likely to be a much longer-term project, FSD Beta, Tesla’s technology that is pushing its vehicles toward full self-driving.
Here’s what you need to know about AI Day.
How can I watch Tesla’s AI Day?
The company has yet to share details about the AI tuning event, but it will likely be streamed on its YouTube account as the first Tesla AI Day. That’s what Musk used for major events for the automaker and two of his other companies, rocket maker SpaceX and brain-computer interface designer Neuralink.
The exact timing is not yet clear, but in-person participation begins at 5 p.m. PT Friday, per event invitation.
What is Tesla up to with the Tesla Bot?
When Musk introduced the idea of the Tesla Bot, he said it would be “friendly” and designed to handle “dangerous, repetitive, boring tasks” that humans would want to avoid. “In the future, physical work will be a choice. If you want to do it, you can, but you won’t have to do it,” he said.
As designed, the 125-pound, battery-powered, human-sized robot will move using 40 mechanical actuators, including 12 in its hands, to perform at a “human level.” It will lift 150 pounds and perceive the world with eight cameras. Tesla’s design demonstrated this by using the same computer that drives Tesla cars.
It is extremely difficult to build a robot that can handle the various environments that humans inhabit. But operating in a more confined and controlled environment, such as a Tesla warehouse or factory, reduces complications. This is like if Tesla started working on an autonomous car by making the Autopilot feature work only on highways where the lanes are well marked, there are no traffic lights, and parked cars and pedestrians are rare.
It’s harder to make a humanoid robot than a stocky machine on wheels. Boston Dynamics’ Atlas robot is a good example of how many years research can take. But if you can get one to work, it’s conveniently configured for navigating and manipulating objects in a human-made world.
Expect an Optimus prototype at Tesla’s AI Day.
What exactly is AI?
Artificial intelligence today generally refers to the use of vast amounts of real-world data to train computer systems to recognize patterns, understand what’s going on, and make decisions. It’s a profound change from the narrow strictures of traditional if-this-then-that programming, focusing instead on the ability to grapple with a wider range of tasks that are far more complex and nuanced.
Big Tech is investing billions of dollars in AI for projects like Google search results, Apple iPhone 14 photos, and Facebook’s system for selecting ads based on the text of our posts. It is making gradual progress, although it is still largely used for specific tasks and lacks the universal capabilities of the human brain.
Musk helped found a lab called OpenAI, which is an advanced natural language processing with an AI model called GPT-3, and demonstrated creativity by turning text prompts into works of art with another AI model called DALL-E. OpenAI’s mission is to “ensure that artificial general intelligence benefits all of humanity.”
What is Tesla doing with AI?
Tesla is a major player in the field of AI. Its FSD Beta software is one example, but the Tesla Bot that roams our homes, responds to our commands, or moves boxes of screws around Tesla’s factory floor will also need to use AI.
One of the most difficult parts of artificial intelligence is training the model, an effort that requires data center computing power. Sophisticated AI models can take days or weeks to train.
Tesla created a technology called Dojo to speed up AI training focused on receiving video data from Tesla cars. The basis is its custom-designed AI processors, which can be connected by the thousands into a single “exapod”. Expect more on Dojo at AI Day.
For the input video to make sense, objects in it, such as bicycles, left turn arrows at traffic lights, and stop signs, must be labeled. It’s partly done by humans and partly by other AI systems, and Tesla has invested heavily in auto-tagging.
Once the AI model is trained, other Tesla-designed processors in the company’s cars run the model to identify its surroundings and make decisions about how to drive.
Even there, Tesla developed its own AI chips. “Tesla’s inference engine is among the best in performance… which gives them a real advantage,” said Keith McMillen, CEO of BeBop Sensors, a startup that designs touch sensors that robots can use.
What is the status of Tesla’s FSD technology?
FSD, short for Full Self Driving, is a technology that is actually categorized as driver assistance by automotive industry standards. Tesla requires a person to pay attention and always be ready to take control from the computer if necessary, and enforces its policy of requiring the driver to apply some pressure to the steering wheel frequently.
FSD is an advance over Tesla’s assistance technology called Autopilot, which only works on highways. But Tesla struggled for years to deliver FSD to customers who paid for it. In recent months, FSD opened up the Beta program to drivers who earn high safety scores, as judged by monitoring Tesla behavior such as braking hard or following other cars too closely.
In September, Musk released a software update to a much larger customer base — 160,000 FSD beta testers in total, he tweeted. FSD Beta version 10.69.3, due for release in October, “brings incremental improvements,” Musk also said.
Is robot AI harder or easier than car AI?
The same technology Tesla is developing for cars can be adapted for humanoid robots.
“Our cars are semi-sentient robots on wheels — neural networks that recognize the world and understand how to navigate the world,” Musk said at the first AI Day.
But the variety of situations a robot can encounter in a person’s home is much more diverse than what a car will encounter on the road, as well as the tasks we can command it to do. A robot that works in more confined conditions, such as a factory or warehouse, would be easier to train.