OpenAI Scales Robotics Lab To Train Robots
OpenAI has quietly built and scaled a robotics lab in San Francisco since February 2025, employing about 100 contract data collectors to teleoperate Franka robotic arms around the clock. The program focuses on collecting large household-task datasets—tasks like folding laundry and placing toast—using 3-D-printed GELLO controllers and plans a second lab in Richmond, California. The effort signals OpenAI's push into devices and robotics training capabilities.
Key Points
- 1Employs about 100 contract data collectors teleoperating Franka arms to record household-task datasets.
- 2Aims to scale large robotics datasets cheaply, contrasting motion-capture or full-humanoid training approaches.
- 3Enables faster model training and simulation use but highlights continued reliance on human labor.
Scoring Rationale
Provides notable new company-level details and dataset scale, but relies on insider reporting and lacks official confirmation.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems