\n<\/p>\n
Robotics company Boston Dynamics has struck a partnership with Google\u2019s AI research lab to speed up the development of its next-generation humanoid robot Atlas \u2014 and make it act more human around people.<\/p>\n
The partnership, which was announced Monday during the Hyundai press conference at CES 2026<\/a>, is centered on robotics research that will use Google DeepMind\u2019s AI foundation models. Boston Dynamics\u2019 humanoid robot Atlas will be the first test case, according to Carolina Parada, senior director of robotics at Google DeepMind.<\/p>\n \u201cWe\u2019re looking to integrate our cutting-edge AI foundation models with Boston Dynamics\u2019 new Atlas robots, and we\u2019ll aim to develop the world\u2019s most advanced robot foundation model to fulfill the promise of true general-purpose human needs,\u201d Parada said onstage.<\/p>\n The tie-up comes less than a year after the Google AI research lab announced new AI models called Gemini Robotics that are designed to allow robots to perceive, reason, use tools, and interact with humans. Gemini Robotics is based on a large-scale multimodal generative AI model, Gemini. At the time, Google DeepMind said the robotics AI model was trained to generalize behavior across a range of different robotics hardware.<\/p>\n Enter Boston Dynamics, and its majority owner, Hyundai Motor Group. While accelerating research will be a central piece of this partnership, this has real-world scaling intent.<\/p>\n Boston Dynamics already has products, like the quadruped Spot, that are in customers\u2019 hands in more than 40 countries. Its warehouse robot Stretch has unloaded more than 20 million boxes globally since its launch in 2023, according to Hyundai. Now Boston Dynamics and Hyundai are preparing for the next generation, starting with the humanoid robot Atlas, which the company announced Monday is already in production and headed to the Hyundai factory in Savannah, Georgia<\/p>\n A prototype of Atlas walked onstage during the press conference, showing off its ability to move. But as Alberto Rodriguez, director of Atlas behavior at Boston Dynamics, noted, making \u201cAtlas into a product requires more than athletic performance for humanoids to really deliver on their promise. They have to be able to interact with people naturally.\u201d<\/p>\n Techcrunch event<\/p>\n \n\t\t\t\t\t\t\t\t\tSan Francisco<\/span> Rodriguez and his counterparts at Boston Dynamics believe that recent advancements in AI have created a clear path to get to those capabilities. That kind of natural interaction with humans has real safety implications. <\/p>\n The Atlas product, which was also revealed onstage Monday and will eventually head to Hyundai\u2019s factory, has 56 degrees of freedom with rotation joints and human-scale hands that have tactile sensing. And it\u2019s strong. The Atlas robot can lift up to 110 pounds and is designed to perform repetitive movements. <\/p>\n With that kind of dexterity and strength, it will be critical for Atlas, or any humanoid robot, to safely interact and work with humans. Some of that has been handled on the hardware side; Atlas, for instance, has 360-degree cameras to allow it to see when people are approaching. But DeepMind\u2019s work could help the robots learn how to act.<\/p>\n \u201cRather than having a set of predefined, loaded tasks onto the robot, we think robots should understand the physical world the same way we do,\u201dParada said. \u201cThey should be able to learn from their experience. Should be able to generalize new situations and get better over time. So whether it is to assemble a new car part or to tie your shoelaces, robots should learn the same way we do from a handful of examples, and then get better very quickly with a little bit of practice.\u201d<\/p>\n Hyundai, which plans to bring Atlas to its factory this year and eventually deploy them for tasks like parts sequencing by 2028, has also developed protocols to increase safety and efficiency.<\/p>\n Hyundai said Monday it is opening a U.S. facility this year called a Robot Metaplant Application Center, or RMAC, that will teach robots how to map movements like lifts and turns. Training data from RMAC will be combined with real-world data collected via a software platform used in its Georgia factory to continually improve the robots.<\/p>\n Follow along with all of TechCrunch\u2019s coverage of the annual CES conference here<\/a>. <\/p>\n
\n\t\t\t\t\t\t\t\t\t\t\t\t\t|<\/span>
\n\t\t\t\t\t\t\t\t\t\t\t\t\tOctober 13-15, 2026<\/span>\n\t\t\t\t\t\t\t<\/p>\n<\/p><\/div>\n<\/p><\/div>\n<\/div>\n