Become a fan of Slashdot on Facebook

 


Forgot your password?
Close
typodupeerror
×
AI Microsoft Software Games Entertainment Technology

Microsoft's AI Is the First to Reach a Perfect Ms. Pac-Man Score (theverge.com) 59

Posted by BeauHD from the game-over dept.
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.
This discussion has been archived. No new comments can be posted.

Microsoft's AI Is the First to Reach a Perfect Ms. Pac-Man Score More

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

Comments Filter:

  • Dupe (Score:1)

    by Anonymous Coward

    This is a dupe. Slashdot editors suck and should all be fired.

  • Dupe (Score:5, Informative)

    by xororand ( 860319 ) on Thursday June 15, 2017 @02:27AM (#54623999)

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

    • Re: (Score:3)

      by bungo ( 50628 )

      Since it's about Ms Pacman, I think this is a homage to us old timers, who remember when the original Pacman came out, and when Slashdot still had Taco, and we'd have dupes of dupes every day.

      Why, if we didn't have at least two dupes a day, we'd complain!

      This is just be current owners reflecting on the old days.

      Look, someone with a 3 digit id is now going to post telling me to get off his lawn (although I was around before accounts existed and didn't want to register as I didn't like being tracked on the in

    • Re: (Score:1)

      by Anonymous Coward

      And uncovered as a hardwired fraud-

      https://www.theregister.co.uk/2017/06/15/microsoft_pac_man/

    • Yes but we didn't have cellphones then.

  • Why Ms. Pac-man? (Score:5, Informative)

    by DNS-and-BIND ( 461968 ) on Thursday June 15, 2017 @03:13AM (#54624089) Homepage
    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.
  • Comment removed based on user account deletion

  • ...UUUUUUUU... oh wait. Too late.

  • Of course Google just a few weeks ago made a lot of buzz with AlphaGo ; *This* is an amazing achievement. And MS had to catch up! But Ms PM compared to AlphaGo ... well, not comparable.

    • AlphaGo, so far as I can tell, was just Deep Q Learning applied to a different game with more hardware resources. This is a different coding paradigm. I consider that far more interesting.

      • "Just" deep q learning? Vs a more classical resolution scheme... Well, Pac Man is funnier than Go, but the challenge is very different.

        • Well, yeah, just deep q learning. I'm not saying deep q learning wasn't important when deepmind applied it to atari games; I'm saying as far as I can tell alphago took existing tech and applied it to go. Hence, less interesting. Because alphago isn't advancing the state of the art.

          Now, I have no reason to think this is more important than deep q learning....

  • FFS, does no-one on slashdot know how to encode text on the web? Does no-one give stories even the most cursory of proof reading?

Slashdot Top Deals

Every nonzero finite dimensional inner product space has an orthonormal basis. It makes sense, when you don't think about it.

Close