Learning dexterity
OpenAI trained a human-like robot hand to manipulate physical objects with unprecedented dexterity, using reinforcement learning and simulation.
5 articles
OpenAI trained a human-like robot hand to manipulate physical objects with unprecedented dexterity, using reinforcement learning and simulation.
OpenAI publishes variational option discovery algorithms for reinforcement learning. These methods enable agents to automatically learn reusable sub-tasks without explicit supervision, improving efficiency and generalization.
OpenAI launches its first Scholars 2018 program, training experienced software developers in machine learning. The program is now underway and publicly documented.
OpenAI introduces Glow, a reversible generative model using invertible 1×1 convolutions. The model generates high-resolution realistic images, supports efficient sampling, and discovers manipulable features. Code and online visualization tool released.
OpenAI trains an agent to achieve 74,500 on Montezuma's Revenge from a single human demonstration, surpassing all published results. The algorithm replays sequences from key states in the demo and optimizes score using PPO.