Redwood City, California – on a recent morning, knocked on the front door of an elegant two -story house in Redwood City, California. In a matter of seconds, the door was opened by a faceless robot, dressed in a beige jumpsuit that perfectly adjusted to his slender waist and long legs.
This slender humanoid greeted me with what seemed to be a Scandinavian accent, and I offered a hand grip. When our palms met, he said, “I have a firm squeeze.” When the owner of the house, a Norwegian engineer named Bernt Børnich, ordered a bottle of water, the robot turned, walked to the kitchen and opened the fridge with one hand.
Artificial intelligence is already driving cars, writing tests and developing computer codes. Now, humanoids – machines built to look like human and fueled by AI – are about to enter our homes to help with daily tasks. Børnich is CEO and founder of a startup called 1x. Before the end of the year, your company expects to put its robot, Neo, in more than 100 residences in Silicon Valley and elsewhere. Your startup is among the dozens of companies that plan to sell humanoids to homes and businesses.
Since 2015, investors have injected $ 7.2 billion into more than 50 startups, according to Pitchbook, a research company that monitors the technology sector. Humanoid fever hit a new peak last year, when investments exceeded $ 1.6 billion. This value does not include the billions that Elon Musk and Tesla, his electric car company, are investing in Optimus, a humanoid who started building in 2021.
Entrepreneurs like Børnich and Musk believe that humanoids, one day, will do much of the physical work that is currently done by people, including housework such as cleaning counters and emptying washes, work in warehouses such as separating packages and working in factories such as assembling cars in production lines. Simpler robots – like small robotic arms and autonomous trolleys – have long shared the workload in warehouses and factories. Now companies are betting that machines can take on a wider range of tasks by imitating human movements, such as walking, folding, cheering, reaching, holding and generally performing activities.
Videos on the internet have been circulating for years showing the remarkable dexterity of these machines, but they are often guided remotely by humans. And simple tasks, such as carrying the dishwasher, are far from simple to them. “There are many videos out there that give a false impression on these robots,” said Ken Goldberg, a professor of robotics at the University of California in Berkeley. “Although they seem human, they do not always behave like humans.”
Continues after advertising
Neo said “Hello” with a Scandinavian accent because he was being operated by a Norwegian coach in the basement of Bøornich’s house. (In the future, the company plans to create call centers where perhaps dozens of technicians can support robots.) The robot walked alone through the dining room and the kitchen. However, the coach spoke for Neo and remotely guided his hands using a virtual reality headset and two wireless joysticks.
Robots are still learning to navigate the world autonomously. And at least for now, they need a lot of help to do that.
“I saw a hardware level that I didn’t think was possible”
I visited the 1x offices in the Silicon Valley for the first time almost a year ago. When a robot named Eve entered the living room, opening and closing the door, I couldn’t avoid the feeling that this wide -eyed robot was actually a costume person. Eve moved on wheels, not legs. Still, it seemed human. I thought of “The Sleep” (“Sleeper”), the science fiction comedy of Woody Allen, 1973, full of robotic stewards. The company’s engineers had already built Neo, but he had not yet learned to walk. An initial version was hung on the wall of the company’s laboratory.
Continues after advertising
In 2022, Børnich entered a zoom call with an AI researcher named Eric Jang. They had never met. Jang, now 30, worked in a robotics lab at Google headquarters in Silicon Valley, while Børnich, now 42, drove a startup in Norway named Halodi Robotics. A potential investor had asked Jang to gather information about Halodi to assess whether it was worth investing. Børnich introduced Eve, something he dreamed of building from a teenager, inspired – like many roboticists – by science fiction (his personal favorite: the 1982 movie Blade Runner).
Jang was fascinated by the way Eve moved. He compared the call at a science fiction drama scene “Westworld”, where a man participates in a party and is shocked to find that everyone in the room is robots. “I saw a hardware level that I didn’t think was possible,” said Jang.
The potential investor did not invested in Halodi. But Jang soon convinced Børnich to unite forces. Jang was part of a Google team that taught new skills using mathematical systems called neural networks, which allow robots to learn from data that represent real -world tasks. After seeing Eve, Jang told Børnich that they should apply the same technique to humanoids.
Continues after advertising
The result was a transatlantic company that they renamed 1x. The startup, which has grown to about 200 employees, now has more than $ 125 million in investor financing, including Tiger Global and OpenAi.
“All this behavior is learned.”
When I returned to the company’s laboratory about six months after I met Eve, I was received by a neo who already knew how to walk. They taught him to walk entirely in the digital world. By simulating real -world physics in a video game -like environment, they managed to train a digital version of the robot to get up, balance and eventually take steps. After months training this digital robot, they transferred everything he had learned to a physical humanoid.
If I got in the way of Neo, he would stop and divert me. If I pushed his chest, he remained standing. Sometimes he stumbled or didn’t know exactly what to do. But he could walk a room similar to people. “All this behavior is learned,” Jang said while Neo made a click against the floor with each step. “If we put it in any environment, he should know how to do this.”
Continues after advertising
Training a robot to perform housework, however, is a completely different perspective. As the physics involved in carrying a dishwasher or folding clothes is extremely complex, 1x cannot teach these tasks in the virtual world. She needs to collect data within royal houses.
When I visited Børnich’s house a month later, Neo began to have difficulties with the refrigerator’s stainless steel door. The robot’s Wi-Fi connection had fallen. But as soon as the hidden technician restarted Wi-Fi, he guided the robot smoothly in his small task. Neo handed me a bottle of water. I also watched Neo carry a washing machine, crouching carefully to get clothes from a laundry basket. And while Børnich and I talked outside the kitchen, the robot began to clean the counters. All of this was done by remote control.
“What we are selling is more a journey than a destination”
When guiding Neo on household chores, Børnich and his team can collect data – using cameras and other sensors installed on the robot itself – which show how these tasks are performed. Then 1x engineers can use this data to expand and improve Neo’s skills.
Just as ChatgPT can learn how to write academic works by analyzing texts taken from the internet, a robot can learn to clean windows by identifying patterns in digital video hours.
Building a humanoid like Neo costs approximately the same as making a small car – tens of thousands of dollars.
To achieve its potential, Neo needs to capture videos of what happens inside the houses. In some cases, technicians will be able to observe what happens in real time. Essentially, this is a robot you learn while working.
“What we are selling is more of a journey than a destination,” said Bøornich. “It will be a way really full of obstacles, but Neo will do things that are truly useful.”
