Genius AI, a startup that aims to develop a basic model to strengthen all kinds of robots, emerged from the Steel with a $ 105 million seed round associated with Aclaps ventures and Khosla Ventures.
Last December was founded by Zhou Xiyan (the above photo, left), PhD. In Carnegie Melne University’s robotics, and in the French AI lab research scientist, Tapofel Griffet, startup, wants to make a general purpose model that will enable the robot to automate numerous repetitive tasks from lab work to housekeeping.
Large models of language are trained on a wide range of text datases, but the AI ​​model for robotics should be trained on the physical world data. However, getting this real -world data makes expensive and timely effort.
To overcome this, birth is turning to artificial data, which it uses a proprietary physics engine that, which says, is capable of modeling the physical world properly.
Birth artificial data engine began with an educational project that Xiyan in collaboration with researchers from 18 universities. Several participants of the project have initially joined, and has created an existing staff of more than 20 researchers who specialize in robotics, machine learning and graphics.
Birth claims that its proprietary fake engine model is allowed to faster, which is a separate advantage than competitors who rely on NVIDIA software.
Other companies working on developing a general -purpose AI model for robots include physical intelligence, which enhanced a Million 400 million rounds; And the Skill AI, which was worth 4 billion earlier this year.
“This is a huge unknown: Do anyone have a great model of the Robotics Foundation that will normalize the tasks? This is a condition after which we want to go,” said Tech Crunch.
He added, “Of all the teams we have seen, we like the birthplace (birth) to go behind the models of the Robotics Foundation.
Birth is producing its artificial data and is building a basic model in two offices in Silicon Valley and Paris.
As the next milestone, birth plans are planned to release its model in the robotics community by the end of the year.