Another use of artificial intelligence highlights our great abilities to excel in humans, this time with the famous old Pac-Man game in which Microsoft achieved the highest achievable result.
A team from the deep-rooted learning firm, Maluuba, worked on the use of a branch of artificial intelligence called Learning Enhancement to play the version of the Atari 2600 from the game. Pac-Man.
With the “divide-and-conquer” method, the artificial intelligence system can reach as many points as 999,990 points.
Professor Doina Precup at McGill University in Montreal expressed her admiration not only for the achievement of researchers in the artificial intelligence of Microsoft’s emerging company, but also for the way in which they achieved it.
The team used the method of breaking up the big problem into smaller problems and distributing them to several branches to deal with them through artificial intelligence to reach the highest result. It is interesting to work on several different parts to achieve one goal in the end.
The team has adopted a method called Hybrid Reward Architecture, which uses 150 agents to work together in parallel to win the game.
The difficulty of the game is artificial intelligence because it simulates human thinking and intelligence because of the complexity of the game and the limited number of times it plays (spirits).
Finally, may you think of the usefulness of such an achievement in practice? According to the company’s video, this development in artificial intelligence can improve decision-making in difficult situations such as optimal marketing channels, financial modeling and even robotics.