Machines are soon going to take over Humans.
British Artificial Intelligence Agency’s team designed automatic “agents”. In a course of time, these Agents automatically learned about multiplayer video game shooter. In the process of learning, they became experts in it and made a habit to win over humans.
When it comes to one on one AI system, humans can still have authority over it because of it being less complicated. However, it is the multiplayer games where artificial gamers are succeeding these days because it requires intense teamwork and interaction from human games.
After the 1997’s remarkable win of IBM’s Deep blue over Garry Kasparov, Computer players are proving to be better gaming opponent as compared to humans. In 2017, World’s No. 1 Go player lost to Google Artificial Intelligence Agent.
Max Jaderberg’s team studied about the latest version of Quake III Arena with “capture the flag” mode. In this mode, the player needs to catch the opponent team’s flag and save their own at the same time. It was released in 1999 and still is one of the best in eSports. The results were shocking as Human testers against agent team won only 25% of games even after practicing for 12 hours.
New steps for Artificial Intelligence:
“Reinforcement learning “is the programmer’s new favorite to enhance the agent’s smartness. According to Jaderberg’s statement to AFP, earlier the agents knew nothing about the world and used to do random stuff only. But now, Agents are taught to reward themselves on achieving any goal. Moreover, today, a team of agents are trained simultaneously at a time rather than teaching a single agent.
DeepMind has said that AI is never going to be part of the military for ethical issues as well as for national interest. However, Jaderberg’s team is looking forward to playing the complete Quake III Arena.