On May 21, IEEE Spectrum published an article detailing how open-source AI is rapidly gaining traction in the robotics field. Hugging Face launched LeRobot, a community platform for robotics AI, in May 2024; since then, the number of hosted datasets has surged from 1,145 at the end of 2024 to over 58,000 today, making it the largest individual dataset category on the Hugging Face Hub. Additionally, citing the fact that software alone isn’t sufficient, Hugging Face announced its acquisition of robotics firm Pollen Robotics to expand into hardware development. Clément Delangue, CEO of Hugging Face, shared the article and expressed his agreement, joking that the accompanying illustration depicted the company’s logo in an “evil version.”
Over the past two years, Hugging Face, NVIDIA, and Alibaba have all invested heavily in the open-source robotics space. The goal of related tools and models is to tackle the far more challenging task of enabling robots to “reason, make decisions, and act” — essentially “making robots start thinking” — rather than merely handling basic motor control. All of NVIDIA’s open-source robotics models are hosted on Hugging Face. According to the article, while this shift toward applying the open-source AI movement to embodied intelligence remains in its early stages, platforms like ROS and LeRobot are gradually making sophisticated robotics AI technologies — previously accessible only within top-tier research labs — available to a broader developer community. This trajectory closely mirrors how the open-source wave accelerated the widespread adoption of LLMs in earlier years.