At Tesla’s AI Day event in August, CEO Elon Musk promised the company would build a humanoid robot called Tesla Bot (code name “Optimus”), and that we’d likely see a prototype next year.
A lot has happened to Tesla in 2021. As the year draws to a close, I’d like to revisit the Tesla Bot with some time removed to break down what has to be Musk’s wildest claim to date.
Also this year, Tesla released the Plaid version of the Model S, which accelerates to 60 miles per hour in two seconds and hits 200 mph — making it the quickest accelerating production car in the history of the world. The electric-vehicle maker is now producing cars at a 1 million annual run rate, and is building more factories than ever before. Tesla is on fire, and that includes its stock price.
Back to the Bot. First, the basics. The 5-foot, 8-inch robot is designed to weigh 125 pounds, be able to carry 45 pounds and move as fast as 5 miles per hour. It will have five fingers on each hand and perform “boring, repetitive and dangerous” tasks. To build it, Tesla TSLA, -3.50% will rely on a myriad of sensors, joints, actuators, the company’s self-driving technology implemented in its cars and advances made in the development of the AI.
Tesla sympathizers and influencers came out of the woodwork to speculate on the brave new world the company will usher in come 2022. The same individuals (and some media outlets) were equally ecstatic in 2016 when Tesla announced Solar Roof on the “Desperate Housewives” set.
Fast forward to 2021, and the company is facing significant issues with what turned out to be a rushed, inadequately researched product, causing cost overruns for Tesla, and delays and headaches for customers.
Five years ago, Musk also promised full self-driving, or FSD, a feature far removed from Level 5 autonomy, the highest. In 2021, FSD is still at Level 2, as Tesla was forced to disclose in a letter to California’s Department of Motor Vehicles.
The (broken) promise of robotaxis
The trend never seems to end. Musk had a similar slip of the tongue in 2019, when he announced 1 million autonomous Tesla robotaxis would be on the road by 2020. Instead of cars capable of autonomously shuttling passengers around, making money for their owners, we got “smart summon,” a feature that’s anything but autonomous.
Right after the rollout, owners complained about various issues, ranging from weird pathfinding patterns to collisions resulting in thousands of dollars of damage. Even this hit-and-miss feature was oversold by Musk.
But this is just a tip of the iceberg. The self-proclaimed “Technoking of Tesla” made so many promises he never fulfilled that there are now websites dedicated to his failures to deliver.
Overselling and underdelivering
Tesla Bot seems to be the case of same-old, same-old. Musk is still doing what he does best — overselling and underdelivering. “Our cars are semi-sentient robots on wheels,” he has said. No, they’re not. They can barely pull off the smart summon feature, let alone demonstrate any kind of sentience other than limited Level 2 autonomy provided by FSD.
In the case of the Tesla Bot, provided Musk ever builds one, these inadequacies will be even more evident. The eyes of the robot will be the same cameras used for Tesla’s Autopilot feature, and a “full self-driving” computer will be placed in the torso. So, that’s the same FSD that requires a driver’s constant attention. It’s also the one that causes collisions and makes wrong calls.
Finally, a vehicle’s movement is far simpler than that of a robot. Robots, especially humanoid ones, need to execute far more complex maneuvers using inverse kinematics, servos and actuators, as well as plethora of sensors in order to make sense of, and to interact with, the environment.
It’s unreasonable to expect that a navigation system that is still in development and proved to be inadequate even for use in cars would somehow significantly evolve by 2022 in order to master complex environments Tesla Bot is expected to navigate.
Those bots should be able to autonomously buy groceries. Judging by the current performance of the FSD, you can expect them to bump into people, get hit by cars, get lost as they miscalculate trajectories, fall over and get damaged, and so on. I don’t believe they could pose a threat to humanity other than getting in the way.
More complex than a car
Simply getting a robot to walk straight and navigate terrain is a gargantuan feat in itself. Boston Dynamics and Agility Robotics have been trying to crack this issue for many years. They’ve achieved impressive results in the end, but it’s been a long and bumpy ride that only serves to demonstrate the enormity of the problem. Musk could be able to solve it, but judging from his track record, that won’t happen by next year.
Does this mean the Tesla Bot will never see the light of day? Not necessarily. Musk usually tries hard to accomplish the projects he promotes and is a much better engineer than a promoter. Indeed, if he were less hype-y about the Tesla Bot announcement, this article would have a totally different tone.
Chances are that the event served to push the stock price higher, sidestep the fear, uncertainty and doubt caused by investigative news, appease investors and attract new talent for Tesla’s robotics division. Musk has been a trailblazer for many industries, and here we see him doing the same for robotics. Consumer and industrial robotics have been around for years, but never like this.
As Musk stumbles over himself to make a Tesla Bot prototype, crashing through many deadlines and prototypes in the process, he may disappoint many of his ardent followers, but I’m certain he will also solve multiple genuine problems in robotics, which will ultimately serve to push forward the idea he was trying to promote with his robot.
A more realistic timeframe
In the process, Tesla may end up being the first company that creates robots such as the Tesla Bot but highly unlikely in the timeframe Musk proposes.
So, what timeframe could be more realistic?
In an interview for Reuters, Raj Rajkumar, a professor of electrical and computer engineering at Carnegie Mellon University, was asked that same question. He said he was fairly certain that “it will be much longer than 10 years before a humanoid bot from any company on the planet can go to the store and get groceries for you.”
Do you think Tesla has what it takes to pull this off? Finally, what’s your take on robots like the Tesla Bot? Do we really need them? Let me know in the comment section below.
Add Comment