Elon Musk is known for his outlandish predictions, but even by his standards, the forecast stood out.
Last year, the Tesla chief executive said the number of two-legged robots on earth would one day surpass the 8 billion humans on the planet.
“I think we might exceed a 1:1 ratio of humanoid robots to humans,” Musk told shareholders at Tesla’s investor day last March, predicting that the arrival of the machines would dramatically upend the global economy.
Despite Tesla being the world’s most valuable carmaker, Musk predicted that the company’s robotics work would be “worth significantly more than the car side of things, long term”.
It is a bold statement. In 2021, When Musk first announced plans to produce a 170cm-tall droid designed to do “boring, repetitious and dangerous” work, he did not even have a prototype. Instead, he brought out an actor in a bodysuit and helmet, who danced for a few seconds before shuffling off the stage.
It was a neat illustration of the gap between the promise and the reality of robots.
Humanoids – bipedal machines that mimic the human body – have been a science fiction staple since the genre’s creation, and an engineering ambition since Renaissance Italy (Leonardo da Vinci’s notebooks contain several drawings of mechanical knights).
In the 1980s and 1990s, the Japanese company Honda ploughed hundreds of millions into ASIMO, a humanoid that largely became known for viral videos of it failing to master stairs.
However, the robots we have today are typically single-purpose machines, such as the arms on robot production lines or the spinning platforms in Amazon warehouses that resemble large Roombas.
Humanoid robots that can sense their environments, walk on two legs and grasp objects in their hands – requirements that could see them replace humans in manual jobs – would be a potential game changer but progress has remained elusive.
Robots have repeatedly struggled to master tasks that seem elementary to humans. Take walking – humans learn to totter around on two legs with moderate success at around 12 months, but the task has turned out to be a fiendishly difficult science problem for robots.
Darpa, the US military’s big-spending research lab, ran a robotics challenge in 2015 in response to the Fukushima nuclear disaster, with the aim of producing a robot that could handle the sort of dangerous environments that humans would wish to avoid.
The challenge pitted robots against a rudimentary obstacle course of door handles and stairs; the majority failed spectacularly, unable to stay on their feet.
The recent boom in artificial intelligence, however, has revived hopes that we will soon live in a world full of droids that will clean our houses, carry out dangerous tasks – and potentially take our jobs. Developments in artificial intelligence systems such as ChatGPT are translating into robots that better understand the world.
In 2022, a year after Musk had announced his robot plans with a man in a suit, he unveiled a real robot, one that could walk, wave, and carry boxes. Musk has said the robot, called Optimus, would cost around $20,000 (£15,800) and said Tesla planned to start using it in its own factories before a public release.
Optimus is not the only humanoid making rapid progress.
In February, researchers at the University of California, Berkeley, trained a two-legged robot to walk around San Francisco. The researchers said they had trained the robot on YouTube videos of humans walking, using similar systems to how AI software such as ChatGPT are developed. This gave the robot a greater understanding of the world around it and allowed it to “generalise”, walking on terrains and roads that it had never encountered without having to be shown it beforehand.
Jensen Huang, Nvidia’s chief executive, has suggested that these types of advances will lead to progress in robotics dramatically accelerating. Once robots are capable of learning by watching humans move, their progress should be rapid, since footage of people is among the most plentiful data that is out there.
Last month, the AI chip giant Nvidia unveiled a suite of software called Project GR00T, designed to help humanoid robots learn about the world around them.
“While the hardware for these robots evolved rapidly over the years, the software lagged behind,” says Deepu Talla, the vice president of robotics and edge computing at Nvidia. “Recent advancements in simulation technology, generative AI and edge computing are closing this gap. Robots are evolving into self-learning machines.”
The prospect of useful humanoids being around the corner has led to an investment boom. In February, Nvidia, Microsoft and OpenAI, the company behind ChatGPT, invested $675m in Figure, a two-year-old Silicon Valley start-up that aims to put robots into warehouses.
Amazon has invested in Agility Robotics, another humanoid company. It is already testing Agility’s 5 foot 9 inch bots, which feature glowing white eyes and paddles at the end of their arms, in its warehouses where they move empty boxes. It is a rudimentary task but one that there is a huge market for, says Pras Velagapudi, Agility’s chief architect.
“We’re already in markets that can support thousands or tens of thousands of robots,” he says. Velagapudi says the next step is expanding its robots’ ability to manipulate objects so it can handle more warehouse jobs.
Agility says its goal is for buying a robot to pay off in two years. Its machines will soon be rolling – or walking – off a production line that is scheduled to be operational by the end of the year.
China, meanwhile, is pushing ahead with robotics technology as a national priority. Beijing’s Ministry of Industry and Information Technology has said it wants to be building humanoids by next year.
However, Ash Sharma, the co-founder of Interact Analysis, which advises companies on how to bring automation technologies into their supply chains, says manufacturers are not rushing to buy humanoid robots.
“It is not a proven use case right now. There are some test cases, some pilots that are happening, but these are not being widely used.” Sharma calls some of the funding going into the industry “eyewatering” given that “No one has said ‘This is the holy grail. This is what we need’.”
That might be a relief to today’s factory and warehouse workers. Agility has estimated the cost of “employing” its robot could be around $3 an hour – well below minimum wage. Widespread and cheap humanoids would be likely to set off a new wave of unemployment fears.
Sharma believes those concerns are probably overblown. “There is a severe shortage of labour in factories and warehouses. And companies are not able to hire people quickly enough. Companies like Amazon have made it very clear that they are lacking labour. They’ve had to push up wages.”
Musk does not agree. In a discussion with Rishi Sunak last year, he told the Prime Minister that “there will come a point where no job is needed… the AI will be able to do everything”.
While it may be yet more Musk hyperbole, the world he has imagined is less outlandish than it seemed just a year ago.
Disclaimer: The copyright of this article belongs to the original author. Reposting this article is solely for the purpose of information dissemination and does not constitute any investment advice. If there is any infringement, please contact us immediately. We will make corrections or deletions as necessary. Thank you.
