Robot Foundation Models Expected to Shift Toward General-Purpose Physical Intelligence by 2035

A chapter from the STATERA Library's Future AI Native Robotics series examines how robot foundation models may evolve between 2025 and 2035. It describes a move away from separately engineered perception, planning, and control modules toward shared representations that link sensing, language, and physical action. Early vision-language-action models are expected to dominate the 2025 to 2027 period, though they face limits from scarce physical interaction data and inconsistent hardware. Later development may favor reusable skills and cross-platform training, with data diversity and training methods mattering more than raw model size. The authors also point to internet-scale video and human activity data as a complementary source for learning object relationships and task structure.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)
Log in to join the discussion and vote.
Log in