Boston Dynamics joins forces with its former CEO to speed up the study of its Humanoid Robot Atlas

Boston Dynamics Wednesday announced a partnership Designed to bring improved strengthening of training for his Electric Humanoid Robot Atlas. The connection is with Robotics Institute and AI (RAI Institute)Early known as the AI ​​Institute of Boston Dynamics.

Both organizations were founded by Mark Ribert, a former professor of MIT, who has been Boston Dynamics CEO for 30 years. The Institute, founded in 2022, allows Raibert to continue research that served as the basis for Boston’s dynamics.

Both have relationships with Hyundai. Korean carmaker acquired Boston dynamics in 2021; Hyundai also funded the institute, giving Raibert Free Rein to explore more expensive and bleeding technologies than possible in a trading company. The Institute reflects the creation of Tri by Toyota, for the Toyota Research Institute, which announce your own partnership With Boston Dynamics in October focuses on the use of models of large behaviors. (Lbms).

Twin partnerships are designed to improve the way Boston Dynamics Electric Atlas Humanoid learns new tasks. The deal for the Robotics & AI Institute is specifically focused on strengthening the training, a method that works through experiences and mistakes similar to the way both humans and animals learn. Strengthening training is traditionally extremely intense, although the creation of an effective simulation has allowed many processes to be carried out at a virtual setting.

Boston Dynamics/Rai Institute Union launches earlier this month in Massachusetts. This is the most in a number of cooperations between the couple, including joint efforts to develop a research training kit for Quadrupal Spot Robot from Boston Dynamics (which is his known dog “Dog”). The new work focuses on the transfer of simulation -based training, to the real -world settings and to improve the way the company’s humanoid atlas moves and interacts with the physical environment.

As for the latter, Boston Dynamics points to “dynamic running and manipulating the whole body of heavy objects”. Both are examples of actions that require the synchronization of the legs and arms. The Humanoid’s bipdal factor presents a number of unique challenges – and opportunities – compared to Spot. Every activity is also the subject of a wide range of forces, including balance, strength, resistance and movement.

A larger picture, Raibert notes in a statement: “Our goal for RAI is to develop a technology that allows for future generations of intelligent machines. Working on Atlas with Boston Dynamics allows us to make progress in strengthening training for perhaps the most complicated humanoid robot. This work will play a decisive role in progressing the capabilities of humanoids not only by expanding its skill set, but also for optimizing the process to achieve new skills. “

The partnership news arrives the day after founder of AI figure and CEO Bret Adcock declared that Bay Area abandons partnership with Openai in favor of developing own internal models.

“We have found that in order to resolve the embodied AI on a scale in the real world, you have to integrate vertically AI into the robot,” the CEO told TechCrunch. “We cannot assign AI for the same reason that we cannot assign our hardware.”

The figure has determined that the best AI models for his humanoid are those that are designed specifically for his robots, internal ones. Openai’s approach to embodied intelligence – meaning in physical form – is understandably less focused, given the size and scope of the Chatgpt manufacturer. This news also arrived on the heels of rumors that OpenAi examines the creation of his own humanoid robotsS

Most companies involved in humanoid space work on their own AI models. This is certainly true of Boston Dynamics, which has decades of experience in developing software to work in its own unique robotics systems. While the RAI Institute is technically a separate organization, the two share the parent company, founder and, probably common goals.