AI
Universe
OpenAI has released Universe, a software platform designed to test and train artificial intelligence (AI) in various tasks. The platform allows develo...
AI
AI
#Exploration: A study of count-based exploration for deep reinforcement learning
Researchers at OpenAI have published a study on using count-based methods for exploration in deep reinforcement learning. This approach involves keepi...
AI
OpenAI and Microsoft
OpenAI is partnering with Microsoft to run its large-scale AI experiments on the latter's cloud platform, Azure. This move allows OpenAI to scale up i...
AI
On the quantitative analysis of decoder-based generative models
Researchers at OpenAI have published a paper on analyzing and evaluating decoder-based generative models. These models use a neural network to generat...
AI
A connection between generative adversarial networks, inverse reinforcement learning, and energy-based models
Researchers have discovered a connection between three types of AI models: generative adversarial networks (GANs), inverse reinforcement learning, and...
AI
RL²: Fast reinforcement learning via slow reinforcement learning
Researchers at OpenAI have developed a new algorithm called RL², which accelerates the process of training reinforcement learning models. By using a s...
AI
Variational lossy autoencoder
A variational lossy autoencoder is a type of neural network that can compress and reconstruct data while learning its underlying distribution. This mo...
AI
Extensions and limitations of the neural GPU
OpenAI researchers have explored the capabilities and limitations of a neural GPU architecture, which combines traditional computing with artificial n...
AI
Semi-supervised knowledge transfer for deep learning from private training data
Researchers at OpenAI have proposed a method for semi-supervised knowledge transfer in deep learning models. This approach allows private training dat...
AI
Report from the self-organizing conference
A conference on machine learning was held at OpenAI's offices, with over 150 attendees. The event focused on self-organizing systems, which are a type...
AI
Transfer from simulation to real world through learning deep inverse dynamics model
Researchers at OpenAI have developed a method for training AI models to learn the underlying dynamics of simulated environments and apply that knowled...
AI
Infrastructure for deep learning
Deep learning research relies heavily on high-quality infrastructure, which can significantly impact progress in the field. The good news is that the ...