\n<\/p>\n
Jay Li doesn\u2019t recommend getting sued by Tesla if you\u2019re trying to get a startup off the ground. But he does think his company, Proception, might be better off for having endured the experience. <\/p>\n
\u201cI think it\u2019s kind of like a resilience test, or pressure test,\u201d he told TechCrunch in an exclusive interview. \u201cPeople say that what doesn\u2019t kill you makes you stronger, right?\u201d <\/p>\n
Li, who was a technical lead on Tesla\u2019s Optimus humanoid robot program, was accused by his former employer last year<\/a> of absconding with trade secrets to start Proception. But after months of trading<\/a> legal blows, he finally reached a settlement with Tesla, which dismissed the lawsuit earlier this month. (Tesla did not respond to a request for comment.)<\/p>\n Now Li is free to tackle what he thinks is an even harder problem: making robot hands work like a human\u2019s. <\/p>\n To help 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. <\/p>\n Proception also announced Monday that it is shipping the first batch of its \u201chigh-dexterity robotic hand\u201d to \u201cresearchers and robotics companies,\u201d while opening up to wider orders. The goal, Li said, is to become the top hand supplier to other companies that don\u2019t want to spend the time or resources developing what\u2019s known in the industry as \u201cdextrous manipulation.\u201d<\/p>\n While there\u2019s been an avalanche of money and attention rushing into the world of robotics, Li believes not enough of that has gone to making robotic hands truly mimic a human\u2019s hands.<\/p>\n One of the loudest voices talking about this challenge has actually been his old boss, Tesla CEO Elon Musk, who has said robot hands are one of the biggest engineering problems yet to be solved. <\/p>\n While Musk has maintained that Optimus robots could start working in factories in a matter of years, the consensus view is that making robotic hands equivalent to a human\u2019s is still many years away. Kevin Lynch, the director of Northwestern University\u2019s Center for Robotics and Biosystems, told the Wall Sreet Journal last year that his team believes it will be a decade until they are \u201cfunctional and useful and able to do some of the things that humans do.\u201d<\/p>\n Li thinks Proception can do it much faster, in large part because of how they\u2019re collecting data. <\/p>\n Most companies training humanoid robots right now are using teleoperators to train their systems. A human wearing a virtual reality headset is able to see what a robot sees and manipulate what\u2019s in front of that robot, then the robot can learn from the commands given by the human.<\/p>\n A big drawback to this approach, according to Li, is that the teleoperator is not receiving feedback from the objects the robot is touching. This approach is also limited to the number of robots a company has available at any given moment, Li said.<\/p>\n Proception\u2019s solution is a glove laden with sensors. With human testers wearing the gloves (and a headset), Proception and its customers can capture \u201chuman hand interaction data without requiring a robot in the loop,\u201d according to Proception\u2019s press release.<\/p>\n This same glove also goes on the hand Proception is developing, acting as its sensor-packed \u201cskin.\u201d The hand has 22 degrees of freedom and multiple joints per finger to enable a \u201cwide range of dexterous motions,\u201d according to Proception.<\/p>\n Li said this approach will also let Proception and its customers gather finer, more task-specific data that can allow its robotic hands to more accurately resemble a human\u2019s. He also thinks it is better suited to scale up.<\/p>\n \u201cYou need both hardware and data, and those need to come hand-in-hand to get [dextrous manipulation] to work. A lot of companies solely focus on hardware, or like hardware plus non-scalable data [collection],\u201d he said. \u201cWe\u2019re working on this highly dexterous hardware plus highly scalable data. We believe that\u2019s a key combination to solve this problem.\u201d<\/p>\n First Round partner Bill Trenchard, who led the investment in Proception, said this was a big reason why he backed Li. <\/p>\n \u201cWe think they will have the best hand in the market, maybe the most sophisticated hand today, and the underlying data and models to support that,\u201d he told TechCrunch. \u201cDexterous manipulation is a very, very, very important part of the whole humanoid story going forward, and as many people have said, it\u2019s sort of the last mile of getting these robots to be truly performant.\u201d <\/p>\n Trenchard also praised Li\u2019s ability to keep a cool head while being sued by his former employer. <\/p>\n \u201cHe was very upfront with us when this came out, and I think the team did an amazing job of keeping their heads down,\u201d Trenchard said. \u201cJay\u2019s a very strong leader.\u201d <\/p>\n Li is also confident. After facing down Tesla\u2019s \u201chardcore litigation department<\/a>,\u201d he told TechCrunch that he wouldn\u2019t be surprised if the company comes calling for help as Proception grows. <\/p>\n \u201cI think it will happen,\u201d he said. <\/p>\n<\/div>\n When you purchase through links in our articles, we may earn a small commission<\/a>. This doesn\u2019t affect our editorial independence.<\/em><\/p>\n