Scalpel Meets Silicon: AI Trains Robots to Master Surgery from Video Alone

A Mitsubishi Electric robotic arm efficiently organizes components, showcasing advanced automation in a modern industrial setting.

Researchers from Johns Hopkins and Stanford have successfully trained a surgical robot to perform complex tasks—such as suturing, manipulating needles, and lifting tissue—by using videos and the same machine learning architecture that powers ChatGPT. Traditionally, robotic surgical systems like da Vinci require precise, code-based instructions and years of development for specific procedures. However, this new approach uses video input to teach the robot relative movements, enabling greater adaptability and faster learning. For example, if a robot drops a needle mid-surgery, it can autonomously recover and continue the task without human intervention.

The system was trained using footage from cameras mounted on da Vinci robotic arms, originally meant for post-op analysis. By combining this video data with kinematic modeling—mathematical representations of motion—the researchers were able to teach the robot surgical skills without direct programming. The team is now aiming to train a robot to perform entire surgeries autonomously. If successful, this innovation could make robotic surgery more efficient, accessible, and reduce reliance on highly specialized human operators.