Jay Li doesn’t recommend getting sued by Tesla if you’re trying to start a startup. But he believes his company, Proception, could be better for having endured the experience.
“I think it’s kind of like a stress test or a stress test,” he told TechCrunch in an exclusive interview. “People say what doesn’t kill you makes you stronger, right?”
Lee, who was technical lead on Tesla’s Optimus humanoid robot program, was he was accused by his former employer last year of absconding with trade secrets to launch Proception. But after months marketing legal blows, eventually reached a settlement with Tesla, which dropped the lawsuit earlier this month. (Tesla did not respond to a request for comment.)
Now Lee is free to tackle what he believes is an even more difficult problem: making the robot’s hands work like human hands.
To do that, Proception announced Monday that it has raised an $11 million seed round led by First Round Capital, with contributions from Y Combinator and early-stage fund BoxGroup.
Proception also announced Monday that it is sending the first batch of its “high-dexterity robotic arm” to “robotics researchers and companies” while opening up to wider orders. The goal, Li said, is to become the go-to supplier to other companies that don’t want to spend time or resources developing what’s known in the industry as “skillful handling.”
While there has been an avalanche of money and attention rushed into the world of robotics, Li believes that not enough has been done to make robotic hands truly mimic human hands.
One of the loudest voices talking about this challenge was actually his old boss, Tesla CEO Elon Musk, who said robot hands are one of the biggest engineering problems yet to be solved.
While Musk has claimed that Optimus robots could start working in factories within a few years, the consensus is that becoming human-equivalent robotic arms is still years away. Kevin Lynch, director of Northwestern University’s Center for Robotics and Biosystems, told the Wall Street Journal last year that his team believes it will be a decade before they are “functional and useful and able to do some of the things that humans do.”
Li believes Proception can do this much faster, in large part because of the way it collects data.
Most companies that train humanoid robots currently use remote operators to train their systems. A human wearing a virtual reality headset is able to see what a robot sees and manipulate what is in front of that robot, then the robot can learn from the commands given by the human.
A big drawback of this approach, according to Li, is that the remote controller does not receive feedback from the objects the robot is touching. This approach is also limited by the number of robots a company has at any given time, Li said.
Proception’s solution is a glove loaded with sensors. With human testers wearing the gloves (and a headset), Proception and its customers can capture “human hand interaction data without requiring a robot in the loop,” according to Proception’s press release.
This same glove also goes on the hand Proception grows, acting as her sensor-filled “skin.” The hand has 22 degrees of freedom and multiple joints per finger to allow for a “large range of dexterous movements,” according to Proception.
Li said this approach will also allow Proception and its customers to gather finer, more specific data on tasks that can allow its robotic hands to more accurately resemble human hands. He also believes that it is more suitable for scaling.
“You need both hardware and data and they have to come hand in hand to get them [dextrous manipulation] to work Many companies focus exclusively on hardware or like hardware and non-scalable data [collection]”, he said. “We’re working on this highly dexterous hardware and highly scalable data. We believe this is a key combination to solve this problem.”
First Round partner Bill Trenchard, who led the investment in Proception, said that was a big reason he backed Li.
“We think they will have the best arm on the market, probably the most sophisticated arm today, and the underlying data and models to support it,” he told TechCrunch. “Dextrous manipulation is a very, very, very important part of the whole humanoid story going forward, and as many people have said, it’s the last mile to make these robots really perform.”
Trenchard also praised Li’s ability to keep his cool while being sued by his former employer.
“He was very biased towards us when this came out and I think the team did an amazing job of keeping their heads down,” Trenchard said. “Jay is a very strong leader.”
Lee is also confident. After facing ‘Tesla’hard-line judiciary,” he told TechCrunch that he wouldn’t be surprised if the company comes asking for help as Proception grows.
“I think it will happen,” he said.
When you purchase through links in our articles, we may earn a small commission. This does not affect our editorial independence.
