Microsoft's AI Is the First to Reach a Perfect Ms. Pac-Man Score (theverge.com)

Maluuba, a deep-learning team acquired by Microsoft in January, has created an AI system that has achieved the perfect score for Ms. Pac-Man. According to The Verge, the AI system "learned how to reach the game's maximum point value of 999,900 on Atari 2600, using a unique combination of reinforcement learning with a divide-and-conquer method." From the report: Though AI has conquered a wealth of retro games, Ms. Pac-Man has remained elusive for years, due to the game's intentional lack of predictability. Turns out it's a toughie for humans as well. Many have tried to reach Ms. Pac-Man's top score, only coming as close as 266,330 on the Atari 2600 version. The game's elusive 999,900 number though, has so far only been achieved by mortals via cheats. Maluuba was able to use AI to beat the game by tasking out responsibilities, breaking it up into bite-sized jobs assigned to over 150 agents. The team then taught the AI using what they call Hybrid Reward Architecture -- a combination of reinforcement learning with a divide-and-conquer method. Individual agents were assigned piecemeal tasks -- like finding a specific pellet -- which worked in tandem with other agents to achieve greater goals. Maluuba then designated a top agent (Microsoft likens this to a senior manager at a company) that took suggestions from all the agents in order to inform decisions on where to move Ms. Pac-Man. The best results came when individual agents "acted very egotistically" and the top agent focused on what was best for the overall team, taking into account not only how many agents wanted to go in a particular direction, but the importance of that direction.

Comments

Dupe

[xororand]

This was already posted only hours ago.
https://games.slashdot.org/sto...

Why Ms. Pac-man?

[DNS-and-BIND]

This particular Atari game was one of the few games that resisted to Deep Q Learning (a form of Reinforcement Learning invented by DeepMind). Many researchers have tried over the last couple of years to solve it. This time, Microsoft found an ingenious solution to the problem, that combines experience from multiple agents and learns to form sub-goals. Their solution could mean that in the future it might be easier to apply reinforcement learning to other settings, such as robotics. The interesting part about reinforcement learning is that it learns dynamic behavior, as opposed to static classification. It learns to act in a way that mimics intelligence. This kind of machine learning is invaluable.