Embodied AI emerges as a grand challenge in the new AI100 report

A new long-term research report examining artificial intelligence has named embodied intelligence and physical AI as a major scientific grand challenge. The study, released by an international committee of experts, assesses how technology has evolved across technical, economic, and societal areas. It compares current advancements with findings from five years prior.

The State of Robotics and Global Leadership

Recent years have brought significant progress in connecting software systems to physical machines. Researchers have combined language processing with robotic controls and improved virtual simulations that let models train safely before moving to real hardware. According to the report, China currently leads the development of industrial robots by deploying more units than the rest of the world combined. The country also manufactures about 80 percent of global humanoid robots, a rapidly growing sector that currently makes up a small portion of the overall market.

Overcoming Real-World Complexity

Despite these hardware and software advancements, experts point out a major gap between lab demonstrations and actual deployment. Many current successes happen only in controlled spaces like warehouses or involve very specific tasks like floor cleaning. The unpredictable nature of the real world creates hurdles that require complex planning and careful handling of varied objects. While some developers believe current training methods are enough to solve these problems, others argue that completely new scientific approaches will be needed to bring advanced robots into everyday environments.

• The study is the third edition in a series that evaluates artificial intelligence every five years.
• China builds the vast majority of humanoid robots and deploys more industrial units than all other nations combined.
• Real-world deployment remains difficult due to the complex nature of physical environments.