DeepMind, owned by Google first built an AI-powered system that was capable of defeating Go and chess and now it is working on games that require multiplayer.
Researchers of DeepMind, in a paper published by Science Magazine, described the way capture flag mode in Quake III was grasped in by the AI system they designed.
450,000 rounds were played of capture the flag to train the AI agents, which can be in a human’s team or maybe against them. AI agents are players in the game that took a few weeks to train the gameplay of four years.
Like any AI system, initially it was unable to achieve the target but after a little training, AI agent succeeded in developing a winning strategy.
One of the researchers at DeepMind, Wojciech Czarnecki said that arbitrary skills allow AI agents to acclimate to teammates.
The paper mentioned that the AI agent won against humans even after their response time was slowed down, to match that of humans. Still, the win-rate had been more than ever before, against humans or even agents.
Warehouse robots that are expected to work in a group can utilize the abilities of DeepMind’s AI, said report in The New York Times. Self-driving cars can benefit from this as well to handle heavy traffic.
Mark Riedle, a Computing professor at Georgia Tech College said that the AI agents of systems are unable to interconnect with one another and rather only react to what is happening in the game. This means that DeepMind needs to improve its existing system. The company might use this multiplayer technology to advance the efficiency of its AI.
Read next: Artificial intelligence can bring life in Mona Lisa's portrait
Researchers of DeepMind, in a paper published by Science Magazine, described the way capture flag mode in Quake III was grasped in by the AI system they designed.
450,000 rounds were played of capture the flag to train the AI agents, which can be in a human’s team or maybe against them. AI agents are players in the game that took a few weeks to train the gameplay of four years.
Like any AI system, initially it was unable to achieve the target but after a little training, AI agent succeeded in developing a winning strategy.
One of the researchers at DeepMind, Wojciech Czarnecki said that arbitrary skills allow AI agents to acclimate to teammates.
The paper mentioned that the AI agent won against humans even after their response time was slowed down, to match that of humans. Still, the win-rate had been more than ever before, against humans or even agents.
Warehouse robots that are expected to work in a group can utilize the abilities of DeepMind’s AI, said report in The New York Times. Self-driving cars can benefit from this as well to handle heavy traffic.
Mark Riedle, a Computing professor at Georgia Tech College said that the AI agents of systems are unable to interconnect with one another and rather only react to what is happening in the game. This means that DeepMind needs to improve its existing system. The company might use this multiplayer technology to advance the efficiency of its AI.
Read next: Artificial intelligence can bring life in Mona Lisa's portrait
