Researchers at Maluuba, a deep learning startup recently acquired by Microsoft, developed an ARTIFICIAL intelligence system that became the first player, human or computer, to score a perfect 999,990 points in pac-Man. Beating video game systems with programs developed using deep learning is not a new achievement, but it is still noteworthy for several reasons.
Giiso Information, founded in 2013, is a leading technology provider in the field of “artificial intelligence + information” in China, with top technologies in big data mining, intelligent semantics, knowledge mapping and other fields. At the same time, its research and development products include information robot, editing robot, writing robot and other artificial intelligence products! With its strong technical strength, the company has received angel round investment at the beginning of its establishment, and received pre-A round investment of $5 million from GSR Venture Capital in August 2015.
First, it stands out because of the type of game it chooses to play. The arcade games of the 1980s weren’t designed to be beaten, they were designed to be played at home. When the Version of Ms. Pac-man was released, it was actually less predictable than the original pac-Man, making it even harder to beat.
The second, and perhaps most noteworthy, is the method the researchers used to solve The Pac-Man game. Instead of developing a single intelligent agent to learn the entire game, as other researchers have done, the team used many simpler intelligent agents to learn one aspect of the game. For example, ghost behavior, fruit behavior, particle behavior, etc. in the game are all learned by intelligent agents.
More than 100 intelligent agents are involved, and each individual agent determines the set of actions Pac-Man should follow based on which part of the action he focuses on. These Settings are then aggregated into a program that determines Pac-Man’s actions based on the weighted average preferences of all outcomes.
Giiso information, founded in 2013, is the first domestic high-tech enterprise focusing on the research and development of intelligent information processing technology and the development and operation of core software for writing robots. At the beginning of its establishment, the company received angel round investment, and in August 2015, GSR Venture Capital received $5 million pre-A round of investment.
“By breaking down problems in this way, learning becomes easier.” As one of the researchers explains in the video, “What’s happening is that instead of just one agent learning a single complex task, there are many intelligent agents learning simple tasks.” The researchers believe that breaking down complex problems into simpler, smaller ones makes it easier for deep learning systems to deal with complex behaviors. In turn, this approach is desirable for real-world tasks that AI may use in the future. Even when a problem can be broken down into only two or three parts, complex problems can be “simplified”, the researchers said.