AI
Solving Rubik’s Cube with a robot hand
Researchers have used neural networks to train a robot hand to solve the Rubik's Cube. The system was trained entirely in simulation using reinforceme...
AI
AI
OpenAI Scholars 2020: Applications open
OpenAI is accepting applications for its third class of OpenAI Scholars. The program provides funding, mentorship, and resources to researchers workin...
AI
Fine-tuning GPT-2 from human preferences
Researchers at OpenAI fine-tuned a large language model, GPT-2, using human feedback for various tasks. They found that the model learned to prioritiz...
AI
Emergent tool use from multi-agent interaction
Researchers at OpenAI have observed artificial agents developing increasingly sophisticated tool use while playing a game of hide-and-seek. In a simul...
AI
Evolution Strategies
Evolution Strategies is a type of black-box optimization algorithm that can be used to optimize deep learning models without knowing their precise ana...
AI
Testing robustness against unforeseen adversaries
Researchers at OpenAI have developed a method to test how well neural network classifiers can defend against unexpected attacks. They've created a new...
AI
GPT-2: 6-month follow-up
OpenAI has released its GPT-2 language model, which has 774 million parameters. This comes after previous releases of smaller models and research into...
AI
Learning Day
OpenAI's employees have a unique tradition called 'Learning Day' every Thursday. On this day, they can choose to learn new technical skills that are n...
AI
Microsoft invests in and partners with OpenAI to support us building beneficial AGI
Microsoft is investing $1 billion in OpenAI to support the development of artificial general intelligence (AGI). The partnership aims to create a hard...
AI
Why responsible AI development needs cooperation on safety
A policy research paper from OpenAI identifies four strategies for improving industry cooperation on AI safety norms: communicating risks and benefits...
AI
Meta Reinforcement Learning
Meta-learning is a type of machine learning where the model learns to learn from limited data. In traditional reinforcement learning, agents are train...
AI
OpenAI Robotics Symposium 2019
OpenAI held its first robotics symposium in April 2019. The event brought together researchers and experts to discuss advancements in robotics and art...