Published on : Jun 27, 2018
Valve Corporation has developed a video game named ‘Dota 2’. It is a free to play multiplayer online battle arena (MOBA) game. It is a sequel for Defense of the Ancients, a mod created by a community for Warcraft III- Reign of Chaos for Blizzard Entertainment. The world championship of Dota 2 is forming newer squads for better competition. It will be opposed by a group of artificial intelligence bots. This is because last year, Open AI co-founder Elon Musk took top Dota 2 players while competing with the bots. Now a team of top professionals will be taking part in an exposition match, The International, one of the largest event in eSports.
Last year at The International, Open AI took individual players one-on-one in a minigame match. Pro Dota 2 players said that they were able to learn from the bots, after rewatching the match. Though they came across certain complexities while playing as a team with bots. AI still have to be taught how to coordinate the five bots. Where the bots have to take effective decisions for 10,000s of values, while they also have to process the current situation in the game. Therefore it can be considered more complex than the deep blue supercomputer- a chess game played in 90’s.
To make the bots learn what to do in situations like this, reinforcement learning patterns are used by Open AI. Though it is a trial and error approach, but with time the AI will evolve completely and attain focused style to play in the game. Every day Dota 2 is been run over more than 100,000 CPU’s and AI plays around 180 years’ worth of game. Thus creating an ample amount of learning space for the AI’s to correct themselves as compared to humans. But AI learns in a different manner than humans, therefore the comparison is not justified.