Anthropic announced that its model Claude AI has helped researchers without experience in robotics to program quadruped robots in about half the time required by colleagues who worked without AI support.

The results come from Project Fetch, an internal experiment published on November 12, in which eight researchers were divided into two teams to program Unitree Go2 robots to identify and approach beach balls. The team that used Claude completed the tasks much faster, highlighting the model's ability to reduce obstacles in tasks that slow down work in robotics. The experiment also showed significant differences in team behavior. Participants without AI reported higher levels of frustration, while the team assisted by Claude collaborated effectively. However, questions arose regarding safety, as a recent study showed that AI models can approve commands that could cause harm in the real world.

Anthropic emphasized that current AI systems still lack the necessary intelligence to operate robots independently, but future improvements to the model could accelerate their influence on the physical environment.