AI
Improving language understanding with unsupervised learning
Researchers at OpenAI claim to have achieved state-of-the-art results on various language tasks using a scalable system that combines transformers wit...
AI
AI
GamePad: A learning environment for theorem proving
GamePad is a learning environment developed by OpenAI to teach artificial intelligence systems how to prove mathematical theorems. It's designed to he...
AI
OpenAI Fellows Fall 2018
OpenAI is seeking applicants for its OpenAI Fellows program, a six-month paid apprenticeship in artificial intelligence research. The program allows p...
AI
Gym Retro
Gym Retro is an open-source platform for reinforcement learning research on games. It has expanded from around 70 Atari and 30 Sega games to over 1,00...
AI
AI and compute
OpenAI has released an analysis showing that the amount of computer power used in the largest artificial intelligence training runs has been increasin...
AI
Implementing Deep Reinforcement Learning Models with Tensorflow + OpenAI Gym
The author provides a tutorial on implementing deep reinforcement learning models using TensorFlow and OpenAI Gym. The tutorial builds on previous pos...
AI
AI safety via debate
OpenAI is exploring a new approach to AI safety that involves training agents to engage in debates with each other. The goal is for the agents to lear...
AI
Evolved Policy Gradients
Researchers have developed Evolved Policy Gradients (EPG), a metalearning approach that evolves the loss function of learning agents. This method enab...
AI
Gotta Learn Fast: A new benchmark for generalization in RL
Researchers at OpenAI have introduced a new benchmark to evaluate the ability of reinforcement learning (RL) agents to generalize across different tas...
AI
Policy Gradient Algorithms
Policy gradient algorithms are a type of reinforcement learning method used to train policies that map states to actions. The article provides an over...
AI
Retro Contest
OpenAI has announced a retro contest, a transfer learning competition that evaluates the ability of reinforcement learning algorithms to apply knowled...
AI
Variance reduction for policy gradient with action-dependent factorized baselines
Researchers at OpenAI have developed a new method to reduce variance in policy gradient algorithms. This improvement is achieved by using action-depen...