6 articles
OpenAI releases a multi-goal reinforcement learning benchmark for robotics. The platform provides complex environments and calls on the research community to develop generalizable solutions for diverse robotic tasks.
OpenAI releases eight simulated robotics environments and a Hindsight Experience Replay baseline implementation developed for internal research. These tools enabled training models that work on physical robots. OpenAI also issues a call for robotics research contributions.
OpenAI hosts a hackathon on March 3rd in San Francisco's Mission District featuring talks and development sessions at its office.
OpenAI releases a paper co-authored with Future of Humanity Institute, CSER, CNAS and EFF forecasting malicious AI misuse scenarios and mitigation strategies. Outcome of one year of sustained collaborative research.
OpenAI presents an interpretable machine learning method where AIs teach each other using examples understandable to humans. The approach automatically selects the most informative examples to convey a concept, experimentally validated for both AI and human learning.
OpenAI built a system for entity disambiguation using a neural network to classify words into ~100 automatically-discovered types (non-exclusive categories).