Connectionists: Competition: General Video Game AI

[Tom Schaul]

tl;dr:
Build a general-purpose learning/planning/search agent that can play well
on many games, such as Space Invaders or Zelda -- and compete! Start here:
http://www.gvgai.net/


-------------

The first edition of the General Video Game AI Competition (GVG-AI) will be
held this year in in association with the Computational Intelligence in
Games conference (CIG-2014):
http://www.gvgai.net

The objective is to develop controllers that can play not just a single
video game, but any number of unseen games (within the class of 2D
arcade-style games). The starter kit includes 10 games (with 5 levels
each), for example: Boulder Dash, Missile Command, Sokoban, Space Invaders,
and Zelda. But the general agent will end up being evaluated on 10 unseen
games!

Deadline: August 1st 2014.

A starter package includes some simple algorithms to get off the ground
quickly, including a random agent, an MCTS-based agent, and a
genetic-algorithm based one:
http://www.gvgai.net/sampleControllers.php

The CIG track for competition papers is still open too, here:
http://cig2014.de/doku.php/call_for_papers

Good luck!

Diego Perez, Spyridon Samothrakis, Julian Togelius, Tom Schaul and Simon
Lucas

-- 
http://schaul.site44.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20140520/c8c7fa0c/attachment.html>