Maluuba, the deep learning research startup that Microsoft acquired earlier this year, has made an important breakthrough by developing an AI-based system to achieve the maximum possible score at the classic arcade game, Ms. Pac-Man. The researchers used a branch of AI called reinforcement learning to play the Atari 2600 version of the game to achieve the maximum possible score of 999,990.
The team employed a method they call Hybrid Reward Architecture, which used over 150 agents working in parallel with each other. The agents had their jobs divided with some looking for pellets, while others kept an eye out for the ghosts. All these agents reported to a "super agent" who took their suggestions and used them to decide where to move Ms. Pac-Man.
As to why they picked an arcade game from the 1980's, Rahul Mehrotra, a program manager at Maluuba, explained:
“A lot of companies working on AI use games to build intelligent algorithms because there’s a lot of human-like intelligence capabilities that you need to beat the games,”
In the paper about the accomplishment, author Harm van Seijen said that the breakthrough "enables us to make further progress in solving these really complex problems." Mehrotra added that this method could potentially help a company's sales department by targeting its clients with multiple agents in order to maximize their chances of closing the deal.