Boston Dynamics is supercharging Atlas humanoid robot with AI skills
Boston Dynamics and the Toyota Research Institute (TRI) have announced a joint project to bring AI-based robotic intelligence to the electric Atlas humanoid robot. The partnership will leverage TRI's expertise in large behavior models (LBMs), a technology that is similar to the large language models (LLMs) that power platforms like ChatGPT.
TRI's breakthrough in robotic learning
During last year's Disrupt conference, Gill Pratt, head of TRI, revealed their lab has achieved 90% accuracy in robots performing household tasks such as flipping pancakes. This was done through overnight training sessions. Pratt emphasized the efficiency of this approach by saying, "In machine learning, up until quite recently there was a tradeoff... Now it seems that we need dozens."
Boston Dynamics and TRI partnership: A strategic alliance
While Boston Dynamics is known for its hardware expertise, its contributions in the field of software and AI are also noteworthy. With TRI, the Spot-maker is looking to address the challenges of teaching robots complex tasks. "There has never been a more exciting time for the robotics industry," Boston Dynamics CEO Robert Playter said. He added the partnership brings together two R&D-focused companies to solve real-world problems.
The electric Atlas: A showcase of robotic strength
Boston Dynamics had unveiled its design for the electric Atlas in April, replacing the larger hydraulic model. Then, in August, a brief video of the robot performing pushups was shared, showcasing its impressive strength. The collaboration between Boston Dynamics and TRI is particularly interesting as both organizations are backed by Hyundai and Toyota, respectively - direct rivals in the automotive industry.
The pursuit of a general-purpose machine
The end goal of all these advancements is to build a true general-purpose machine, one that can learn and do everything a human can. While robot hardware has come a long way to this level of sophistication, the challenge of achieving artificial general intelligence is still a long way off. Even with SDKs for systems like Boston Dynamics's Spot, true artificial general intelligence remains an elusive goal.
