AI and games
Jakob
PhD and Postdoc retreat 2021
Department of Physiology, Bern
Montreux, Switzerland, 10.09.2021
[Togelius 2018]
Game AI areas
[Yannakakis & Togelius 2018]
AI plays games to win:
In the beginning there was chess*
"[...] A machine will always remain a machine, that is to say a tool to help the player work and prepare. Never shall I be beaten by a machine! Never will a program be invented which surpasses human intelligence. And when I say intelligence, I also mean intuition and imagination. Can you see a machine writing a novel or poetry? [...]"
Garry Kasparov (1989)
8 years later (1997):
Kasparov looses to "Deep Blue"
How do you play chess?
- imagine moves
- recognize strong positions
white to move
Imagining moves (but with computers): tree search
Initial board
Good board for me after x moves
many, many boards
Pick move that leads towards desired board configuration*
Game AI areas
[Yannakakis & Togelius 2018]
AI models players
"[T]he [computer] modeling of a player’s experience or behavior based on theoretical frameworks about player experience and/or data derived from the interaction of the player with a game."
Why?
- understand player behavior in your game
- understand player experience/feelings
Yeah, but why?
- create believable NPCs
- evaluating game content without requiring playtesters
- personalization of games
- (future) game design
[Yannakakis & Togelius 2018]
Modeling player behaviour
"completionists"
"pacifists", "speedrunners"?
- do people play the game as intended?
- useful to evaluate variations of game design
[Yannakakis & Togelius 2018]
Modeling player experience
"Specifically, fusing the gameplay and the visual reaction features as inputs to the artificial neural network we achieve average accuracies of up to 84%, 86% and 84% for predicting reported engagement, frustration and challenge, respectively." [Yannakakis & Togelius 2018]
[Yannakakis & Togelius 2018]
Game AI areas
[Yannakakis & Togelius 2018]
AI generates game content:
the insatiable hunger for new levels*
Generate new content autonomously
(with or without player models)
[Yannakakis & Togelius 2018]
Who needs humans? AI as game designer: Yavalath
Rules
The board starts empty.
Two players take turns adding a piece of their colour to an empty cell.
Win by making a line-of-4 (or more) pieces of your colour.
Lose by making a line-of-3 pieces of your colour beforehand.
Draw if the board otherwise fills up.
[http://cambolbro.com/games/yavalath/]
Game AI areas
[Yannakakis & Togelius 2018]
Further reading/resources
Togelius, J. (2019). Playing smart: On games, intelligence, and artificial intelligence. MIT Press.
Yannakakis, G. N., & Togelius, J. (2018). Artificial intelligence and games. Springer.
AI and games
By jakobj
