Per reporter: Zheng Yuhang Per editor: Lan Suying
On September 30, local time, Tesla Artificial Intelligence Day (AI Day) opened in Palo Alto, California, USA.
This is the second AI Day event in Tesla's history. Tesla deliberately postponed this year's AI Day by more than a month, mainly to show the public that it can operate the humanoid robot "Optimus".
In that regard, Tesla CEO Elon Musk has indeed said what he said.
At the event site, the prototype of Optimus Prime, under the expectation of everyone, stepped onto the stage, and its internal structure was unobstructed, and the joints, bones, lines, etc. were clearly visible. It waved to the audience and waved its arms in the air. Musk said this is the first time a robot has operated without a tether. In the video shown on site, Optimus can also water flowers with watering cans in the office, handle materials with both hands, and more.
But these capabilities of robots have not impressed industry experts.
A commentator article from the American Society of Electrical and Electronics Engineers (IEEE) bluntly said that Tesla failed to show anything unique and impressive with its new humanoid robot prototype.
Image source: Tesla AI Day event video screenshot
IEEE: Mass production and price expectations "pointless"
The IEEE article said that at the end of Tesla's AI Day 2021 last August, Musk proposed the concept of "Tesla robots", a mechatronically powered autonomous bipedal "universal" humanoid robot. Musk said that the prototype of the Tesla robot will be completed within 2022. At AI Day 2022, the Tesla robot prototype did make an appearance, but it turned out that the previous hype was excessive.
"So far, we've (still) seen little indication that Tesla will be better at solving this problem than anyone else." Tesla doesn't seem to have anything special or exciting that would provide any unique foundation for Musk's vision to surpass other companies doing similar work. I'll reiterate what I said a year ago: the hard part is not building robots, but getting robots to do useful things. The article reads.
For the walking demonstration of the Optimus Prime's prototype on the spot that day, the author of the IEEE article commented that he was not particularly impressed, but only a little confused. At the booth, Optimus Prime's movements were minimal, and his steps were still a little hesitant and stiff. This means that the "actions" it does are likely to be fully scripted, so it is not known to what extent the robot can balance itself.
In fact, Musk also introduced at the time that in order to stay safe, they don't want the robot to do too many actions on the stage and let it "drop a dog and eat."
However, for the ability to water flowers shown in live videos, the IEEE article points out that grasping the watering can is somewhat impressive because it makes it not so easy for the robot to grasp the slender watering can handle. Also, for the robot's action of picking up an object and trying to put it in the trash, but because the video was cut before it was successfully placed, the IEEE said it made them think that the video was a well-planned scheme by Tesla.
At the AI Day site, Musk said that Tesla is confident of achieving mass production at a lower cost, and the final production will reach millions of units. According to Musk, Optimus Prime's price will be less than $20,000 (about 140,000 yuan).
In this regard, the IEEE bluntly said that at present, these "expectations" and figures are meaningless.
TechCrunch also commented that unless Tesla's AI team makes very rapid progress in getting robots to perform tasks like humans, it seems unlikely that a consumer-grade product will launch next year.
Peter J. Bentley, professor emeritus of computer science at University College London and an expert in science media at the Royal Society, commented in an email interview with the Daily Economic News that as long as the funds invested are sufficient, it is not impossible to mass-produce humanoid robots, but whether the robot can complete the work as designed is another matter.
Image source: Tesla AI Day event video screenshot
Expert: There is no need to develop feet, you should focus on the research and development of hands
For the future development of Optimus, the IEEE article believes that in the short and medium term, it is difficult for Tesla to find the positioning of Optimus Prime's unique use.
Alexander Kernbaum, interim director of robotics at SRI International, told CNBC that to create a robot that could function in the factory, there was really no need for Tesla to develop bipedal robots. "For a factory environment, mobility should be as simple as possible, which means walking on wheels, not legs," he explains.
On the one hand, the robot's legs require a lot of power, which will put more load on the batteries that Tesla has developed specifically for robots. In addition, bipedal robots, like humans, are prone to tripped and falling, while wheeled robots are less likely to tip. Kernbaum suggested that safety standards should be benchmarked against factory robots.
Kernbaum believes that Tesla is better suited to focus on the robotic hand. "The hand is like the ultimate multifunctional tool," he says. Dexterity and hand-held object handling is one of the great challenges of 10 years, and it will have a clear impact on all precision manufacturing.
"In fact, how much humanoid robots can do is a composite problem of mobility and artificial intelligence capabilities." Eni Halilaj, an assistant professor of mechanical engineering at Carnegie Mellon University and a researcher who studies human movement, told CNBC that imitating human mobility will be limited, in part because humans currently have limited access to their own knowledge of mobility.
"Our body is a complex engineering system that we still don't fully understand." Khaliraj said. "We still have a long way to go before we can reverse engineer it, which makes motion planning and control of humanoid robots challenging. For example, we still don't understand how our central nervous system chooses specific muscle coordination patterns to perform everyday tasks, which is one of the great challenges of biomechanics and neural control.
"The success of biomimicry depends on our understanding of biological systems." Khaliraj explains.
"A lot of what humans do is very, very difficult for robots, and that doesn't change whether the robot is arm-shaped or human-shaped." Hearst, chief technology officer of humanoid robotics company Agility Robotics, told Reuters.
Daily economic news