AI
From Autoencoder to Beta-VAE
The article discusses the evolution of autoencoders and variational autoencoders (VAEs) in deep learning. Autoencoders were initially designed to reco...
AI
AI
OpenAI Five Benchmark: Results
OpenAI's Five benchmark has achieved a significant milestone by defeating a team of top Dota players in a best-of-three match. The team, consisting of...
AI
Learning dexterity
Researchers have successfully trained a robotic hand to manipulate physical objects with high precision, similar to human dexterity. The achievement i...
AI
Variational option discovery algorithms
Researchers at OpenAI have proposed a new class of algorithms called variational option discovery algorithms. These algorithms aim to improve the effi...
AI
OpenAI Scholars 2018: Meet our Scholars
OpenAI has selected a group of experienced software developers to participate in its OpenAI Scholars program. The first class of scholars is currently...
AI
OpenAI Five Benchmark
The OpenAI Five Benchmark was a series of five matches between the OpenAI Five team's AI and human players in the game Dota 2. The matches were design...
AI
Glow: Better reversible generative models
Glow is a new type of generative model that uses invertible 1x1 convolutions, making it reversible. This allows it to generate high-resolution images ...
AI
Learning Montezuma’s Revenge from a single demonstration
Researchers have developed an AI agent that can achieve a high score on the classic video game Montezuma's Revenge after watching it played only once....
AI
OpenAI Five
OpenAI's team of five neural networks, called OpenAI Five, has begun beating amateur human players in the popular video game Dota 2. The achievement m...
AI
Attention? Attention!
The article discusses attention mechanisms in deep learning, specifically self-attention and its applications. It covers various architectures that us...
AI
Retro Contest: Results
The OpenAI Retro Contest has concluded its first round, where participants were tasked with developing algorithms that can learn and improve over time...
AI
Learning policy representations in multiagent systems
Researchers at OpenAI have developed a method for learning policy representations in multiagent systems. This involves creating a shared representatio...