April 15, 2024

OpenAI-backed 1X just raised $100 million to bring its two-legged butler robot into your home

The idea of ​​two-legged robots tidying up our homes has been the stuff of science fiction for decades. That is now about to change.

Norwegian AI startup 1X today announces its $100 million Series B, led by EQT Ventures.

The company that added OpenAI and Tiger Global to its shareholders for its Series A last year will use the capital to launch its new android NEO, designed for consumers’ everyday household tasks.

From an idea to a bipedal humanoid

1X, formerly known as Halodi Robotics, was founded in 2014 and is based in Moss, south of Oslo. Their first-generation android, called EVE, has been around for four years and weighs 80kg and has wheels instead of legs.

The robot was detected last week. serving coffee to travelers at an Oslo train station.

The company’s mission is to address global labor shortages and its first EVE android is already being used in fields such as logistics and security (surveillance). But with its next-generation Android NEO, 1X is taking on an even more challenging area: people’s homes.

Compared to EVE, NEO weighs only 30 kg, has a soft coating over its aluminum core and has two legs instead of wheels.

1X’s next-gen Android NEO weighs 30kg

Since the robot’s goal will be to help people with their daily tasks at home, the two legs allow it to reach harder-to-reach places, explains 1X CEO and co-founder Bernt Øivind Børnich.

“We don’t mind walking too much. The difficult problem is what we have been working on for many years: manipulation. Having two legs is incredibly important in the home because you need to be able to do so many things in such a small space,” says Børnich, using the example of reaching over the couch to plug something into a wall socket. This would not be possible for a robot on wheels.

Letting a robot work in a warehouse is one thing, but letting it into customers’ homes is a very different challenge, where safety is of great importance. This is also the reason why NEO is clear and smooth-skinned. You will also not take risks such as picking up a heavy object when a person is nearby, in case you drop it.

But teaching robots the kind of generalized understanding of their environment they need to make these decisions is no easy task and one of the reasons they haven’t existed until now.

Have OpenAI and NVIDIA as sparring partners

1X first reached out to ChatGPT developer OpenAI in the summer of 2022, a few months before the iconic chatbot was released to the world. The discussions ended with OpenAI leading the 1X extension of its $23.5 million Series A in early 2023.

“When I started talking to Sam [Altman, CEO of OpenAI] about this, it was before most people thought this [making this type of androids with the help of AI] “It was possible because it was before ChatGPT and there were no real successes to highlight, especially in robotics, which is a few years behind the digital space,” says Børnich.

“Getting that validation for the technology, expanding our network and having the right access to the right type of people and all of this has been really good, but also having a lens to see what’s coming.”

Another company that has played an important role in the growth of 1X is the American technology giant NVIDIA, developer of high-performance GPU cards. In 2023, it published how it uses large language models (LLMs) to automatically generate reward algorithms to train robots to perform complex tasks, and 1X says it can benefit from using NVIDIA’s software on its Androids.

“NVIDIA is more of a partner than a competitor since, at least so far, it is not moving into the hardware space to make its own Androids. It creates the tools that allow us to scale our AI and deploy it, so it’s a very good partnership,” says Børnich.

How does 1X train its androids to perform tasks?

1X is also working on its own custom AI models to control its home robot, a task that requires a lot of training data.

But when you teach robots to put cups in the dishwasher, that data is not something that can be replicated by letting them learn from books on how to become a household help. But according to Børnich, 1X is teaching androids in a similar way to LLMs.

“So the main way to teach robots to do things is by cloning human behavior and thinking into the machine. “This is very similar to how you train large language models,” she says.

Image of 1X's EVE android working in logistics
1X’s android EVE works “happily” in logistics

And while LLMs are trained on large amounts of text, the system that powers 1X’s robot is taught through expert human demonstrations.

“We are gathering the same [as for LLMs] of how it should behave in our world. We have an operator who has a virtual reality headset on, which he sees and hears through the eyes and ears of the robot, and the robot follows the same body movement as the operator,” he says.

“We not only show the behavior of how you perform tasks with your body, but we also try to show the thought process so that the data is labeled with my thought process,” he says.

As the operator says what he does and why he does it at any given time, the different tasks, such as adding cups and plates to a dishwasher and choosing where to put them, will also be added as text to the training system.

“This is how we hope to get AI models that have a much richer understanding of the world. “It is very important for the next steps on the AI ​​path,” he says and that is also why 1X has decided to create a consumer-facing robot. “This is where the huge amount of data and the diversity of data is so we can solve the difficult problems of AI.”

Make robots solve problems they haven’t learned from data

Børnich says advances in LLMs, which can address a wide range of questions and prompts, are transferable to 1X’s AI systems, which should allow its robot to adapt to a wide range of situations without making unsafe decisions. .

“What we do see in our large models is that there is what we call generalization,” he explains.

“This means you understand how to do things that weren’t specifically in your data set depending on the other data you had. And this is incredibly important for us, especially in the home, to be able to have robots that are safe. “Security is something that requires a very, very good understanding of how the world works.”

Today, 1X has 50 robots up and running performing a wide variety of tasks and collecting data in its warehouse in Norway.

“It’s been quite a journey, going from a situation where we had a couple of robots and we were back testing things in the lab, to having 50 robots running 8 a.m. to 4 p.m. every day, including our robot at home, which is ordering after family and learning how that works in the real world,” says Børnich.

The investment

The $100 million Series B was raised so the new Android could be mass produced, and while Børnich isn’t yet ready to say how much it will cost, he says it will be “very affordable” compared to similar systems already out there. The market.

“It hasn’t been long since we raised our Series A, but given the results we’re seeing in our AI stack right now, we see that we’re at the point where we can start scaling these models. And that means expanding our data collection; It means expanding the size of computer data to train our AIs and, of course, it also means expanding our factories and reaching volume manufacturing of our OCTs,” he says.

In addition to EQT Ventures leading the Series B, the round also included a significant secondary transaction with existing Norwegian VC Sandwater increasing its stake with the third largest contribution to the round. Other secondary participants were new investor Samsung Next and existing Norwegian investors Skagerak Capital and the Nistad group.

With the Series B, 1X has raised more than $125 million in total.

Leave a Reply

Your email address will not be published. Required fields are marked *