On August 11, DynaRobotics, an American physical intelligence company founded by three Chinese people, released the next-generation robot basic model DynA-2. The success rate of high-precision manufacturing tasks exceeded 80%. The success rate of performing tasks and the ability to generalize are the two core issues of today's robots. This model, which has been pre-trained based on over 1 million hours of human video, can adapt to the new task of screwing bottle caps using two five-finger dexterous hands in about 13 minutes; moreover, the success rate of high-precision manufacturing tasks increased from 20% of the previous model Dyna-1 to 80%-90%; in scenario deployment, the pass rate of Dyna-2 was 87%, while the pass rate of the previous model Dyana-1 was only 46%.

On August 11, DynaRobotics, an American physical intelligence company founded by three Chinese people, released the next-generation robot basic model DynA-2. The success rate of high-precision manufacturing tasks exceeded 80%. The success rate of performing tasks and the ability to generalize are the two core issues of today's robots. This model, which has been pre-trained based on over 1 million hours of human video, can adapt to the new task of screwing bottle caps using two five-finger dexterous hands in about 13 minutes; moreover, the success rate of high-precision manufacturing tasks increased from 20% of the previous model Dyna-1 to 80%-90%; in scenario deployment, the pass rate of Dyna-2 was 87%, while the pass rate of the previous model Dyana-1 was only 46%.