IEEE Spectrum investigates whether robotics is on the verge of a "ChatGPT moment"—where foundation models trained on broad sensor and motion data enable robots to generalize across tasks without task-specific retraining. Early successes in vision-language models for robotic control and sim-to-real transfer suggest the pattern is emerging.
The bottleneck remains: training data scale (millions of hours of robot interaction), sim-to-real gap in physics engines, and cost of robotic hardware validation. If solved, robotics could mirror AI's recent trajectory from narrow task-specialists to general-purpose agents.