Maluuba, a deep learning company owned by Microsoft, used a type of AI called reinforcement learning to play a perfect game of the Atari-2600 version of Ms. Pac-Man. The final score: the maximum 999,990.
Microsoft reports that while researchers have taught AI to play a number of video games, Ms. Pac-Man has been the most difficult. The team at Maluuba used a method in which more than 150 AI agents worked together on different pieces of the game to achieve a common goal. Because this is similar to how the human brain may work, it could have far-reaching implications for how AI learns new tasks.