Announcement

Collapse
No announcement yet.

Adaptive AI

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Adaptive AI

    During Civ4 there was a Turing test scientist here at Poly, don't remember who it was. He caused the "omgwtfbbq-there's-no-MP-in-vanilla" stir. I think he graduated and got sucked into teaching?

    Fast forwarding to Civ5, the new environment is a steam-centred one, where information between users is easily and commonly exchanged.

    I wonder, why can we not have an adaptive AI? Given that successful strategies can be reverse-engineered into the nuts and bolts that drive them, a computer could match future actions to past results and tinker forth along the way.

    The AI people at Firaxis could work out which general strategies exist, program those, allow the AI to follow on particular paths depending on circumstance, and follow up on settings as they evolve in the field. It's a bit like evolution, only better. It's genetic AI engineering.

    Taking the example of a single player game, the AI's could adapt varying strategies, and future sessions would be based on the strategy that is able to trump the player often. Data on the success and failures against particular competing strategies is uploaded to the server, and reviewed by staff.

    I'm sure this would be a pain to program, but it would be 10 steps closer to ideal. Even deity is like taking candy from a baby right now, and I'd much prefer being beat at prince.


  • #2
    this just in: AI competitions exist for StarCraft 2:

    bgweber writes "The StarCraft AI Competition announced last year has come to a conclusion. The competition received 28 bot submissions from universities and teams all over the world. The winner of the competition was UC Berkeley's submission, which executed a novel mutalisk micromanagement strategy....


    why don't we pitch tailored AI's against each other, to improve the general level of AI?

    Comment

    Working...
    X