Boston Dynamics is teaming up with Toyota to build human-like robots with advanced artificial intelligence (AI). Researchers from both companies will integrate Toyota’s recent advancements in AI with the newest version of Boston Dynamics’ Atlas robots. Atlas robots were originally developed in 2013.
Early models used powerful hydraulic actuators, which posed safety risks to humans. The latest Atlas, unveiled earlier this year, has incorporated electric motors for improved safety. Toyota has developed “large behavior models” for robots, similar to the “large language models” that power AI systems such as ChatGPT.
These models allow robots to learn tasks through remote human instruction without the need for new software. Tasks such as turning the pages of a book, pouring liquids, or operating kitchen tools can be taught through this method.
Integrating AI to advance robotics
The integration of these sophisticated behavior models with a safer Atlas robot offers a competitive edge. It will help Toyota and Boston Dynamics compete with other companies already advancing humanoid robots for various markets. Notably, carmaker Tesla has been showcasing prototypes of its own humanoid robot called Optimus.
Scott Kuindersma, Senior Director of Robotics Research at Boston Dynamics, says “TRI brings significant skills and tools to the table that complement our existing capabilities. This partnership puts us in a great position to achieve some amazing advancements with Atlas.”
Russ Tedrake, Vice President of Robotics Research at TRI, adds “We have a ton of respect for the Boston Dynamics team and what they’ve achieved with Atlas. We want to push the boundaries beyond tabletop manipulation, and this partnership allows us to explore those opportunities.
The collaboration will focus on developing new whole-body manipulation behaviors, collecting performance data to support advanced behavior models, and exploring fundamental training questions for humanoid robots.
This includes understanding the optimal timing for introducing various sensory inputs and actions during training and determining the right mix of tasks to create a robust and dexterous robot. This partnership heralds a significant step towards the practical deployment of humanoid robots capable of operating safely alongside humans in everyday environments. It stands as a key initiative towards achieving general-purpose humanoid robots that can take on a variety of complex tasks in the real world.
Cameron is a highly regarded contributor in the rapidly evolving fields of artificial intelligence (AI) and machine learning. His articles delve into the theoretical underpinnings of AI, the practical applications of machine learning across industries, ethical considerations of autonomous systems, and the societal impacts of these disruptive technologies.
