AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |
Back to Blog
Dota 2 fast push hero3/28/2023 ![]() ![]() Both chess and Go are full-information games. Strong play requires making inferences based on incomplete data, as well as modeling what one’s opponent might be up to. The rest of the map is covered in a fog hiding enemies and their strategies. Partially-observed state. Units and buildings can only see the area around them. ![]() Chess usually ends before 40 moves, Go before 150 moves, with almost every move being strategic. OpenAI Five observes every fourth frame, yielding 20,000 moves. Most actions (like ordering a hero to move to a location) have minor impact individually, but some individual actions like town portal usage can affect the game strategically some strategies can play out over an entire game. Long time horizons. Dota games run at 30 frames per second for an average of 45 minutes, resulting in 80,000 ticks per game.A Dota-playing AI must master the following: The hope is that systems which solve complex video games will be highly general, with applications outside of games.ĭota 2 is a real-time strategy game played between two teams of five players, with each player controlling a character called a “hero”. Relative to previous AI milestones like Chess or Go, complex video games start to capture the messiness and continuous nature of the real world. One AI milestone is to exceed human capabilities in a complex video game like StarCraft or Dota. Follow us on Twitch to view the live broadcast, or request an invite to attend in person! To benchmark our progress, we’ll host a match versus top players on August 5th. This indicates that reinforcement learning can yield long-term planning with large but achievable scale-without fundamental advances, contrary to our own expectations upon starting the project. Using a separate LSTM for each hero and no human data, it learns recognizable strategies. It trains using a scaled-up version of Proximal Policy Optimization running on 256 GPUs and 128,000 CPU cores-a larger-scale version of the system we built to play the much-simpler solo variant of the game last year. OpenAI Five plays 180 years worth of games against itself every day, learning via self-play. We may not succeed: Dota 2 is one of the most popular and complex esports games in the world, with creative and motivated professionals who train year-round to earn part of Dota’s annual $40M prize pool (the largest of any esports game). While today we play with restrictions, we aim to beat a team of top professionals at The International in August subject only to a limited set of heroes. Our team of five neural networks, OpenAI Five, has started to defeat amateur human teams at Dota 2.
0 Comments
Read More
Leave a Reply. |