Tesla Bot is the company’s troubled Autopilot system in humanoid form
Not content to be an automaker or even an energy company, Tesla now wants people to think of it as a robotics company.
At Tesla’s “AI Day” presentation yesterday, CEO Elon Musk made the surprise announcement that the company is working on a humanoid robot. The endeavor, he argued, makes sense given the company’s experience working toward self-driving vehicles.
“Tesla is arguably the world’s biggest robotics company because our cars are semi-sentient robots on wheels,” Musk said. “We think we’ll probably have a prototype sometime next year.”
The robot wasn’t billed as a specialty tool but instead as one that can tackle any manner of challenges. “Can it navigate through the world without having it be explicitly trained, without line-by-line instructions? Can you talk to it and say, ‘please pick up that bolt and attach it to the car with that wrench?’” Musk said. “It should be able to do that. It should be able to go to the store and get me the following groceries, that kind of thing.”
The concept, referred to as “Tesla Bot” and “Optimus,” was shown only in renderings. The company claims the robot will stand 5-feet 8-inches (1.52 m) tall, weigh 125 pounds (56.7 kg), and be powered by 40 electromechanical actuators. Tesla’s slides also called out some impressive-sounding but nebulous features like “human-level hands” and “dojo training.”
“It’s basically got the Autopilot system in it. It’s got cameras [and a] full self-driving computer, making use of all the tools we’ve got in the car,” Musk said.
Unsolved problems
The challenge for Tesla, of course, is that its Full Self-Driving feature has been delayed for years and still isn’t ready for primetime. Betas of the software have been released to a few thousand Tesla owners, and videos of the system in operation show cars driving reasonably much of the time—but making poor choices frequently enough to be concerning. In one example, a car makes a sudden swerve into the right lane before taking a planned left turn. In another, a car attempts to drive into an open hole surrounded by construction cones.
Tesla will face more difficult challenges in developing a robot. While the company currently relies on humans behind the wheel to help train its Full Self-Driving feature—turning the wheel or tapping the brakes when the computer does something wrong—it won’t have the same luxury with a humanoid robot. Some people have the patience to train a car because they have to sit behind the wheel anyway and because of the dire consequences if the computer makes a poor choice. But far fewer people will have the patience to train a robot to perform mundane, low-stakes tasks like selecting an apple at the grocery store, for example. Without thousands or millions of humans voluntarily training Tesla’s robots, the learning process will be exponentially more difficult.
Musk’s timeline of one year to a prototype stage is ambitious, too. Making robots that can navigate complex environments with grace isn’t easy. Just ask Boston Dynamics, which has been working on bipedal and quadrupedal robots for years. Even its advanced machines take nasty tumbles with some regularity.
Musk did allude to using the robots in factories, a setting where automation has obvious advantages. The advantages are so great, in fact, that many factories are already highly automated. In those cases, though, it has been far easier and cheaper to adapt the workspace to suit the needs of the robot than to build a humanoid robot.
Think of the future, man
Musk wrapped up his presentation by waxing philosophical about how robots could undermine the foundation of our economy. For generations, economists, engineers, and philosophers have worried that increasing automation could lead to civilization’s collapse. While that day may come, the concerns have yet to be proven true. Automation has certainly led to job loss and social turmoil, but so far, civilization has been able to adapt.
Experts on the matter frequently point out that automation can either replace workers or augment them, with the latter being preferable. Robots have been used to help surgeons, for example, allowing them to perform new procedures and revise existing ones to be less invasive. But robots can also replace workers, as they have in the automotive industry, which has lost over half a million jobs in the 2000s alone due to automation.
A humanoid robot could threaten many millions of jobs that are currently too challenging to automate. If that happens, then, “essentially, in the future, physical work will be a choice,” Musk said.
How will society adapt if humanoid robots do end up taking our jobs? Musk tossed off a frequently cited solution, universal basic income, saying that it could ease the transition. “But not right now, because the robot doesn’t work,” he said.
https://arstechnica.com/?p=1788840