You didn’t understand Tesla’s economic strategy. This is essentially the message that Elon Musk conveyed during his last quarterly update. Naive as we are, we thought the company’s main product was its increasingly self-driving electric cars. The billionaire, however, revealed that he is betting more on Optimus, Tesla’s mysterious humanoid robot project.
Mentioned for the first time last August, this concept seemed more than embryonic: no prototype had been presented (a dancer disguised as a robot was doing the show…) and the target audience was vague to say the least. We just learned that this robot would measure about 1.70 meters, have a face-screen and be able to “do everything that humans do not want to do” (extensive program). However, Elon Musk is now agitating the idea of producing the first versions of this robot from 2023. “The public has not taken the measure of the scale of the program, affirms the one who is also boss of SpaceX and Neuralink. discerning people will understand that Optimus will eventually be worth more than the car market and more than self-driving. I am sure of that.” So, do we really lack vista?
Robotics has indeed made giant strides in the last ten years and the think tank Idate estimates that it will weigh 90 billion euros in 2030. Advances in artificial intelligence (AI) have given these machines a oh so precious sense: sight. “Robots can now identify where objects are placed that should, for example, be placed in boxes,” explains Jean-Baptiste Mouret, research director specializing in AI and robotics at Inria.
Limited offer. 2 months for 1€ without commitment
Also gone are the days when these companions made of bolts only worked locked in cages so as not to accidentally injure employees. “We now know how to better measure and dose the strength of industrial robots, make them detect contact. They can therefore increasingly operate near humans”, specifies the expert. And in the great race for robotic progress, the regiment of humanoids has won some nice medals.
A stunningly realistic face
After publicizing the pirouettes of its dog robot Spot, Boston Dynamics wowed the public with Atlas, its bipedal robot capable of running, jumping and doing somersaults. The British Engineered Arts, for its part, created the event with its Ameca robot with breathtakingly realistic facial expressions. In his demonstration, the humanoid with his face covered in synthetic skin “wakes up”, looking misty and his eyes flickering, before inspecting his arms and hands with a questioning look, then displaying an expression of deep wonder as he pretends to discover her body. And everything rings incredibly true.
However, Elon Musk seems to forget a little quickly the enormous challenges facing the sector in general, and Tesla in particular. “For the moment, each humanoid robot that exists on the planet has been made ‘by hand’ individually”, explains Justin Carpentier, researcher at Inria specializing in this field. And given the technical challenges that their design entails, this may not change any time soon. If the humanoids now have a steady footing, it pretty much ends there. “They still have trouble managing additional contact points, putting their hand on a wall, a ramp, or holding a fairly heavy object,” explains Jean-Baptiste Mouret. Those who dreamed of having them carry their shopping bags to the supermarket will have to wait…
Tesla arrives very late in the race. Admittedly, Elon Musk’s company has its expertise in computer “vision” on its side: its in-house software that empowers cars is one of the most advanced. But if the eyes are “the window to the soul”, they are only a tiny part of a humanoid robot. A Boston Dynamics thus has a huge lead over Tesla in everything else, especially in the way of making these mechanical bodies move. And even this heavyweight in the sector only manages to “make its incredible achievements in an extremely controlled framework, and certainly not on the first try”, confides specialist Justin Carpentier.
To attract talent to Optimus, the whimsical billionaire took out the checkbook. “They offer some salaries ranging from $200,000 to $300,000,” reveals a robotics researcher. Not sure, however, that this is enough to quickly catch up on the accumulated delay. Last shadow on the board that Elon paints for us (and perhaps the most worrying): the question of the very relevance of the humanoid form for a robot. Of course, creating creations in his effigy is a fad that has been working with humans for ages. But once the “wow” effect of these machines has passed, what is their real added value? The question is crucial, because these robots that look like us cost a fortune.
“The cost of a humanoid robot is very high”
The humanoid form indeed requires many more complex modules than the others (for example, several thousand standardized components and more than 1,000 tailor-made for the Ameca robot). “The difficulty is that the human body relies on a multitude of axes: wrists, elbows, ankles, knees… We must therefore reproduce all of this with mechanical axes and motorized systems”, explains Cyril Kabbara, founder and CEO of Shark Robotics, a French company which, for its part, has opted for more classic robotic forms. As a result, the bill is painful and “the cost of a humanoid robot is generally higher by a factor of 10 than that of a more conventionally shaped robot”, confides an expert in the sector.
To make the customer swallow this pill, it is better to be able to prove that his metal biped brings a lot of added value. And this is where the shoe pinches for manufacturers. Because in many situations humans are, in reality, much more competitive. “No humanoid robot rivals their abilities, recognizes with fair play the CEO of Engineered Arts, Will Jackson. Humans self-repair and self-replicate. They maintain a good state of form for about fifty days. years and are very intelligent.” Even for seemingly simple tasks, this insight makes all the difference. “Robots remain mostly machines programmed once and for all, unable to adapt to an unforeseen situation”, explains Inria research director Jean-Baptiste Mouret.
If they are aesthetically more stunning, humanoids also have trouble competing in many areas with more basic robots. An example ? Their battery life is shorter. A whole part of a robot’s activities (analysis of the environment, etc.) takes place when it is stationary. But for a bipedal device, staying stationary is a completely different challenge than for a robot on wheels or on tracks: the humanoids must activate their motorized systems and perform micro-corrections constantly in order to maintain their precarious balance.
“Operating for an hour remains a challenge for them, while manufacturers often need devices with a minimum autonomy of four hours,” says Cyril Kabbara of Shark Robotics. If a robot jogger throws some in the training room, on the ground it will not always be the most effective. Robots on wheels manage to tow much heavier loads. The tracked machines will play loose ground where the humanoids are likely to sink.
Another problem is that bipeds and quadrupeds cannot go very fast. While Boston Dynamics’ sleek dog Spot runs at 6 kilometers per hour, Shark’s Barakuda crawler travels at 20 kilometers per hour. “A humanoid robot will not be able to lift heavy loads either. This requires stability and an enormous counterweight”, finally highlights Christian Lupsé, expert at the handling equipment dealer Aprolis.
The Express app
To track analysis and decryption wherever you are
Download the app
Ultimately, the situations where the humanoid form provides a decisive advantage are quite limited. For high-risk interventions (nuclear site after an accident, space mission, etc.), this type of robot has an obvious interest. “These devices can sneak into complex and dangerous environments where wheeled robots would not pass and where humans could be injured,” explains Jean-Baptiste Mouret of Inria. But for more prosaic situations, the interest of humanoid robots seems slim. The enthusiastic promises of Elon Musk who sees his Optimus assisting us in the medium term both at the factory and at home therefore seem implausible. But we only ask to be surprised.
Christian Gollier’s chronicle
Detours from France
The chronicle of Jean-Laurent Cassely