Humanoid robots will probably one day be part of our daily lives. If the announcements are multiplying in recent weeks, we are still a long way from functional models as evidenced by the confusing presentation of Tesla last week. But, for CIOs, it is not necessarily too early to take an interest in it…
From C3PO (Star Wars) to R. Daneel Olivaw (The Robot Cycle)… From T-1000 (Terminator) to Roy Batty (Blade Runner)… Humanoid robots have invaded science fiction for a long time. But recent progress in robotics, artificial intelligence and elastic actuators and sensors now makes it possible to create companion robots for professional or family purposes to accompany us on a daily basis in tasks requiring physical strength and the balance.
In recent months, the subject has experienced some media acceleration with the presentations/announcements of Xiaomi’s CyberOne and Tesla’s Optimus. But media times in this area are clearly not aligned with the times needed for R&D. The presentation of “Bumble-C” during Tesla AI Day last weekend was a cruel reminder of this.
Tesla’s Optimus in delicacy…
Last year, Elon Musk caused a sensation by announcing that his company would develop a humanoid robot while presenting to the public a man in a robot costume.
A year later, the first public prototype of the “Optimus” series, called Bumble-C, was able to take its first hesitant steps in front of journalists and observers. A demonstration that disappointed more than one. A disappointment born of the gap that exists between the enthusiastic and unrealistic announcements of Elon Musk and the ambitious technological difficulty. Especially since, for Elon Musk, this presentation was more intended to encourage vocations and recruit talent than to demonstrate Tesla’s progress in humanoid robotics.
Elon Musk is certainly used to the fact but mocking him would be a mistake, because it would be to forget all those who laughed at him in 2005 when he claimed to want to build affordable electric vehicles and make them a mass market or during the two first failures of SpaceX “reusable” rockets in 2006 and 2007.
Bumble-C, Project Optimus’ first public prototype, is nothing short of spooky with its wires and components clearly visible.
So yes, the first prototype of Optimus did not convince anyone. With its mobile lab design far removed from the costume presented last year, its uncertain gait, its undemonstrative swaying, “Bumble-C” is far removed from Tesla’s claims. A deliberately limited demonstration to ensure ” that the robot does not break his face in public “Explained the boss of Tesla who also broadcast videos where his robot did a little more (watering flowers, carrying packages, etc.).
And of course Elon Musk’s hope of seeing this project materialize within three to five years seems hardly credible. ” I think Optimus is going to be amazing, kind of mind blowing, in 5-10 years he admits himself. Many observers consider these deadlines simply unachievable. ” There are many lessons to be learned from understanding how humanoids work “, explains Tom Ryden, executive director of the Mass Robotics incubator. ” But as for having a humanoid directly available to the general public, I doubt that will come out anytime soon. “.
The difficulty of inventing bipedal robots
One of the problems – which illustrates the disconnect between the expectations of observers and the reality of Tesla’s developments – is summarized by Cynthia Yeung, product manager of Plus One Robotics: “ nothing shown is state of the art in the matter.
Elon Musk was well aware of this, recalling that there are more sophisticated robots on the market but which are overpriced and ” brainless “.
The boss of Tesla refers here to the surprisingly agile robots of Boston Dynamics but also to other “bipedal” robots of the early 2000s like the Honda P2, P3, and Asimo, Sony QRio, HRP-2 from Kawada / AIST , M2V2 from Yobotics. He quickly forgets some recent prototypes like the TALOS of PAL Robotics developed with the assistance of the CNRS or the ATLAS from Boston Dynamics.
As Tom Ryden of Mass Robotics explains, ” the hardest part isn’t building a robot, it’s getting that robot to do useful things”
The challenge of a humanoid robot is twofold. First, he must be able to stand up, which is already a feat in itself. Then, it must be able to move independently and perform useful tasks, which requires a good dose of “intelligence” on board.
READ ALSO :
L’Oréal accelerates its R&D with robots
Standing up is much more complex than it seems. This requires an array of sensors of all kinds, complex joint mechanics, and Machine Learning so that the robot maintains its balance at all times and can take advantage of the advantages of being bipedal to walk, climb and descend stairs but also step over obstacles. Moreover, for Tom Ryden, Tesla did not take the easiest and fastest route: “ when trying to develop a robot that is both affordable and useful, a humanoid shape and size is not necessarily the best solution “.
Thus, Tesla still has a lot of progress to make and a lot of work to do so that its robot remains stable, can fall without being damaged and get up on its own. Other robots have certainly done it long before Optimus but have remained impressive demos on YouTube without finding concrete uses.
Because to be useful, these humanoid robots must also acquire enough autonomy to move around on their own and carry out practical daily tasks. That’s what the Tesla boss is implying by “brainless” competitors. But it will also take a lot of effort to invent the AIs of such robots. This is also where Tesla’s interest in developing Optimus lies. According to Elon Musk, “ a robot is like a self-driving car with legs “. If the analogy is disputed by many researchers, Tesla’s side is estimating and developing similar autonomous AIs to drive cars and bipedal robots. Self-driving car research complements Optimus AI research and the reverse is likely true.
READ ALSO :
Cala, the robot that masters the stoves
Still, the autonomy of “circulation” in a universe with unpredictable events is one thing, but making these robots useful is another. As Tom Ryden explains, ” the hardest part isn’t building a robot, it’s getting that robot to do useful things, and I think Musk is still a long way from being able to achieve that goal.”
Laws of robotics at the heart of the debates
In addition to the ability to move safely and act independently, the robot’s AI must also integrate a whole set of safeguards from the outset. Elon Musk explains that the AI of Optimus embeds from the design the concern of ” protect both the robot and the people around the robot while avoiding ‘Terminator-like scenarios’ “.
On Twitter, many observers wondered if and how Tesla had implemented the laws of robotics invented by Asimov in his famous robot cycle (see box below).
Tesla also unveiled the design of the next generation of prototypes. This model was not functional and was brought on stage carried by collaborators.
Be that as it may, Optimus’ presentation has revived debates on social networks about the usefulness of robots and the risks that AI can cause for humanity. A study from New York University also shows that more than one in three AI researchers think that ” AI could cause a global catastrophe on the scale of nuclear war “. The debates on this matter are still far from being settled, but let us remember that the Council of Europe is working on the subject and not only drew up an “AI Act” in 2021 but also published last week two proposals to regulate AI to protect people while enabling innovation.
Tesla and Xiaomi: A focus on mass production
Even if Tesla’s demonstration didn’t live up to the “hype” caused by the Optimus announcement a year ago, two things nevertheless stand out. First in barely a year of R&D, Tesla teams managed to produce a relatively advanced prototype. Then they set out to build a massively “manufacturable” robot.
Competitor “CyberOne” in development at Xiaomi Robotics
Like the Chinese competitor Xiaomi and its humanoid robot Cyber One, Tesla did not design its Bumble-C as a demonstration of the state of the art but as a manufacturable product. ” We designed it using the same discipline as we do in designing our cars, that is, designing it so that it is possible to manufacture the robot in large quantities, at low cost and with great reliability. This is an extremely important and differentiating point: Optimus is designed to be an extremely efficient robot, but manufactured in very large volumes, eventually in millions of units. And it should cost way less than a car – way under US$20,000, in my opinion. explains Elon Musk.
Towards a new civilization?
Build units by the millions… The phrase is not insignificant. This desire to make mass-produced robots will necessarily have multiple impacts in the world of factories, warehouses, hospitals and eventually in homes. The Japanese have long recognized that to cope with an aging population and human service needs, robots will become inevitable. Elon Musk explained in June that “ the importance of Optimus will become evident in the years to come “.
Tesla has unveiled some technical data on the internal components of its prototype.
On the occasion of Tesla AI Day, he developed his vision a little further: There’s still a lot of work to be done to refine Optimus and improve it, and that’s really why we’ve organized this event – to convince some of the most talented people in the world to join Tesla and contribute to its realization, to its realization on a large scale so that it can help millions of people… Optimus allows us to glimpse a future of abundance, a future where there is no poverty, where you can have everything you you want in terms of products and services. This is truly a fundamental transformation of civilization as we know it. “.
In the meantime, if Tesla engineers already manage to get us a reliable and accessible robot that can be used for something, it won’t be so bad…
ISAAC ASIMOV’S THREE LAWS OF ROBOTICS
On the occasion of his cycle of robots, the novelist Isaac Asimov formalized the “laws” that must govern the artificial intelligence of robots so that they can serve humanity usefully and without risk.
First Law : “ A robot may not harm a human being or, remaining passive, leave that human being exposed to danger. »
Second Law : “ A robot must obey orders given by human beings, unless such orders conflict with the First Law. »
Third Law : “ A robot must protect its existence insofar as this protection does not conflict with the First or the Second Law. »
These three laws are encoded in the “hardware” of robotic AI, the positronic brain.
Over the course of the works, the robots, in order not to sink into vicious circles generated by the contradictions of human behavior, will themselves add a ” Zero Law “: ” a robot cannot harm humanity, nor, by its inaction, allow humanity to be harmed “.
We would love to say thanks to the author of this short article for this incredible web content
Humanoid robots: more SF but not yet a reality.
Find here our social media accounts as well as other pages related to it.https://www.ai-magazine.com/related-pages/