We’ve built a system for automatically figuring out which object is meant by a word by having a neural network decide if the word belongs...

We’re releasing a new batch of seven unsolved problems which have come up in the course of our research at OpenAI.

We’re releasing highly-optimized GPU kernels for an underexplored class of neural network architectures: networks with block-sparse weigh...

We’ve developed a hierarchical reinforcement learning algorithm that learns high-level actions useful for solving a range of tasks, allow...

Our latest robotics techniques allow robot controllers, trained entirely in simulation and deployed on physical robots, to react to unpla...

We’ve found that self-play allows simulated AIs to discover physical skills like tackling, ducking, faking, kicking, catching, and diving...

We show that for the task of simulated robot wrestling, a meta-learning agent can learn to quickly defeat a stronger non-meta-learning ag...

We’re releasing an algorithm which accounts for the fact that other agents are learning too, and discovers self-interested yet collaborat...

We’re releasing two new OpenAI Baselines implementations: ACKTR and A2C. A2C is a synchronous, deterministic variant of Asynchronous Adva...

Our Dota 2 result shows that self-play can catapult the performance of machine learning systems from far below human level to superhuman,...

We’ve created a bot which beats the world’s top professionals at 1v1 matches of Dota 2 under standard tournament rules. The bot learned t...

RL-Teacher is an open-source implementation of our interface to train AIs via occasional human feedback rather than hand-crafted reward f...

We’ve found that adding adaptive noise to the parameters of reinforcement learning algorithms frequently boosts performance. This explora...

We’re releasing a new class of reinforcement learning algorithms, Proximal Policy Optimization (PPO), which perform comparably or better ...

We’ve created images that reliably fool neural network classifiers when viewed from varied scales and perspectives. This challenges a cla...