Researchers dispute OpenAI's claim that robot solved Rubik's cube
A robotic hand that OpenAI claimed could solve a Rubik's Cube by itself was heavily criticised by researchers who said the characterisation was highly misleading. Trained using reinforcement learning in simulation, OpenAI's robot hand had a 60 percent success rate for average scrambles and 20 percent for the most difficult scrambles, according to a research paper published by the company. Training the robot hand was done entirely in simulation, with the simulation parameters altered in each training round to make the algorithm more adaptable to different conditions . While considered impressive, the achievement sparked controversy, as OpenAI had used Kociemba's algorithm, a 10+ year old non-learning algorithm to determine the sequence of moves to solve the cube, and yet its video implied the robot learned to solve the cube entirely by itself. Critics compalined that OpenAI had overstated the accomplishment, thereby causing confusion about the system's true capabilities. System 🤖 Dactyl Developer: OpenAI Country: USA Sector: Technology Purpose: Solve Rubik's cube Technology: Robotics; Symbolic algorithm Issue: Accountability; Accuracy/reliability ; Mis/disinformation; Transparency
- Date it happened
- 2019-10-01
- Organisation involved
- OpenAI
- Product, system or model
- Dactyl
This incident was imported from AIAAIC and is used under CC BY-SA 4.0. Our additions to it — the structured fields, the translation, the checks against other reports — are published under the same licence.
This is a record of what was reported, not a finding that anyone broke the law. If it names your organisation and you believe it is wrong, the corrections process is free and open to everyone.