Hacker News
- Better Language Models and Their Implications https://blog.openai.com/better-language-models/ 130 comments
- The International 2018: Results https://blog.openai.com/the-international-2018-results/ 36 comments
- Learning ‘Montezuma’s Revenge’ from a single demonstration https://blog.openai.com/learning-montezumas-revenge-from-a-single-demonstration/ 45 comments
- OpenAI Five https://blog.openai.com/openai-five/ 124 comments
- Improving Language Understanding with Unsupervised Learning https://blog.openai.com/language-unsupervised/ 18 comments
- AI and Compute https://blog.openai.com/ai-and-compute/ 130 comments
- OpenAI Charter https://blog.openai.com/openai-charter/ 177 comments
- Reptile: A Scalable Meta-Learning Algorithm https://blog.openai.com/reptile/ 2 comments
- OpenAI Supporters https://blog.openai.com/openai-supporters/ 38 comments
- Interpretable Machine Learning Through Teaching https://blog.openai.com/interpretable-machine-learning-through-teaching/ 6 comments
- Learning a hierarchy https://blog.openai.com/learning-a-hierarchy/ 39 comments
- OpenAI Baselines https://blog.openai.com/openai-baselines-dqn/ 36 comments
- Roboschool: open-source software for robot simulation https://blog.openai.com/roboschool/ 26 comments
Lobsters
- Better Language Models and Their Implications https://blog.openai.com/better-language-models/ 2 comments ai
- OpenAI Five https://blog.openai.com/openai-five/ 8 comments ai , games
- Robust Adversarial Examples https://blog.openai.com/robust-adversarial-inputs/ 3 comments ai , security
- New AI from OpenAI can generate a story from only 2 sentences of context. https://blog.openai.com/better-language-models/ 15 comments science
- Has anyone seen this yet? https://blog.openai.com/better-language-models/#sample8 5 comments deeplearning
- Big advance in unsupervised language modeling from OpenAI, impressive generated stories https://blog.openai.com/better-language-models/ 3 comments artificial
- Spinning Up in Deep RL - "...an educational resource produced by OpenAI that makes it easier to learn about deep reinforcement learning (deep RL)." https://blog.openai.com/spinning-up-in-deep-rl/ 3 comments learnmachinelearning
- Spinning Up in Deep RL from OpenAI https://blog.openai.com/spinning-up-in-deep-rl/ 3 comments reinforcementlearning
- We've seen Artificial Intelligence conquer Alpha GO, tomorrow at 12:30p PDT; witness OpenAI Bots play against the 99th percentile Dota 2 players. https://blog.openai.com/openai-five-benchmark/ 7 comments technology
- OpenAI DotA update: several restrictions lifted from 5x5 agent games (+wards, +Roshan, fixed hero mirror match ~> 18 heroes), human-equivalent reaction time, just w/more PPO training; pro match at 2PM PST 4 August 2018 https://blog.openai.com/openai-five-benchmark/ 10 comments reinforcementlearning
- "Learning Montezuma's Revenge from a Single Demonstration", Salimans & Chen {OA} [PPO with backward chaining from reward state for curriculum learning] https://blog.openai.com/learning-montezumas-revenge-from-a-single-demonstration/ 3 comments reinforcementlearning
- OpenAI Five [5v5 Dota 2 bots] https://blog.openai.com/openai-five/ 103 comments programming
- OpenAI's new Dota2 Bot beats amateur players in team play https://blog.openai.com/openai-five/ 20 comments artificial
- OpenAI DoTA update: PPO LSTM reaches amateur-level 5x5 DoTA via self-play (256 GPUs/128k CPU, 65000x realtime), highly parallelized framework called 'Rapid'; no hierarchical or abstraction for long-term planning https://blog.openai.com/openai-five/ 17 comments reinforcementlearning
- OpenAI Retro Contest (Sonic meta-RL) results: AliBaba team wins 1st place, 4,692/10,000; 229 submissions; winners use PPO/DQN w/hyperparameter tuning; next contest launches in a few months https://blog.openai.com/first-retro-contest-retrospective/ 4 comments reinforcementlearning
- AI and Compute https://blog.openai.com/ai-and-compute/ 3 comments technology
- Since 2012, the amount of compute used in the largest AI training runs has been increasing exponentially with a 3.5 month-doubling time (by comparison, Moore’s Law had an 18-month doubling period) https://blog.openai.com/ai-and-compute/ 19 comments hardware
- An example of how we all might vote in a world without traditional voting https://blog.openai.com/debate/ 4 comments artificial
- [N] OpenAI: 'Retro Contest' for transfer learning on Sega Genesis _Sonic the Hedgehog_ games (from Steam) w/Gym support as 'Gym Retro' (ends 5 June 2018; trophies promised) https://blog.openai.com/retro-contest/ 10 comments reinforcementlearning
- Why is DeepMinds AlphaZero learning chess/go from scratch a bigger deal than OpenAI learning DOTA from scratch? https://blog.openai.com/dota-2/ 16 comments artificial
- "More on Dota 2" by OpenAI, a blog post about how they trained the bot that beat Dota pros in 1v1 https://blog.openai.com/more-on-dota-2/ 14 comments programming
- Elon Musk's OpenAI first ever to defeat world's best players in competitive DotA 2 https://blog.openai.com/dota-2/ 33 comments worldnews
- OpenAI: human-level 1v1 micro DotA play via self-play deep RL; tournament demonstration https://blog.openai.com/dota-2/ 11 comments reinforcementlearning
- OpenAI: Proximal Policy Optimization variant on TRPO for continuous actions (ALE, Roboschool) https://blog.openai.com/openai-baselines-ppo/ 5 comments reinforcementlearning
- Robust adversarial examples https://blog.openai.com/robust-adversarial-inputs/ 7 comments programming
- OpenAI depevloped evolution strategies as a scalable, faster and easier alternative to Reinforcement Learning https://blog.openai.com/evolution-strategies/ 3 comments technology
- "Evolution Strategies as a Scalable Alternative to Reinforcement Learning" [OpenAI discussion of recent paper, Salimans et al 2017, using neuroevolution for scalable RL] https://blog.openai.com/evolution-strategies/ 3 comments reinforcementlearning