Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Designing Intelligence

Yoeri Staal
@YoeriStaal

info@staalmedia.nl
Adding AI to your game




End of development,
time to put in AI!
How to connect
everything?
Game Theory








A game is a series of
decisions
Info goes in: board
state, options
Decision comes out:
pick a poss...
Game of the Goose


Throw a dice, move
that many steps.



First to the end wins



Not a game!
Rock, Paper, Scissors




Pick any of 3
gestures
Each gesture beats
one other
Turing Test






Impossible to
distinguish human
from computer
AI can pretend to be
a human opponent
AI can replace a ...
Artificial Intelligence








Search Problems
BEST: Find the best
solution from a range
FAST: Find a good
enough sol...
Racing Game


Individual opponents



Multiplayer / Single



Traffic is also AI



Rubber Band
Intelligent Agents






Autonomous entities
Observe through
sensors
Acts upon
environment using
actuators
Intelligent Agents






Autonomous entities
Observe through
sensors
Acts upon
environment using
actuators
Emotions






Change parameters
based on events
Aggressive,
Defensive, Panic,
Revenge...
Class Strategies
Platformer


Player vs World



Run, jump and shoot
Trial and Error








Limited number of
options
Calculate best from
many tries
Pick best or retry
(TAS)
AI can try w...
Real Time Strategy / Sports




Player (human/AI)
controls a team
Each unit is also an
agent
Self-Organizing Systems




Ants leave
pheromones, path to
food
Equivalent to a
shortcut
Shooter




Individual agents
Big difference
monsters and soldiers
Machine Learning




Machine Learning to
study patterns
Might classify player
strategy
Neural Networks


Predict player
behavior



Guard favorite spots



Ambush player
Fighting Game


Player vs Player



Remember combo's



Timing of buttons
Imperfection and Randomness








AI can input combo's
perfectly
AI knows everything
about the world
Add mistakes on...
Dungeon Master






AI controls entire
game world
Weather, characters,
animals..
Even flow of the
game
Conclusion






No single AI solution
Define entities
independent of
human/AI
Find the right strategy
(or mix) for you...
Designing Intelligence

Yoeri Staal
@YoeriStaal

info@staalmedia.nl
Upcoming SlideShare
Loading in …5
×

StaalMedia: Designing Intelligence

800 views

Published on

Spending time in the design and prototyping stage of AI can pay off with a more challenging, more capable and more immersively humanlike game system.

  • Be the first to comment

  • Be the first to like this

StaalMedia: Designing Intelligence

  1. 1. Designing Intelligence Yoeri Staal @YoeriStaal info@staalmedia.nl
  2. 2. Adding AI to your game   End of development, time to put in AI! How to connect everything?
  3. 3. Game Theory     A game is a series of decisions Info goes in: board state, options Decision comes out: pick a possible move Make use of statistics
  4. 4. Game of the Goose  Throw a dice, move that many steps.  First to the end wins  Not a game!
  5. 5. Rock, Paper, Scissors   Pick any of 3 gestures Each gesture beats one other
  6. 6. Turing Test    Impossible to distinguish human from computer AI can pretend to be a human opponent AI can replace a lost player
  7. 7. Artificial Intelligence     Search Problems BEST: Find the best solution from a range FAST: Find a good enough solution fast Making good decision = finding a good solution
  8. 8. Racing Game  Individual opponents  Multiplayer / Single  Traffic is also AI  Rubber Band
  9. 9. Intelligent Agents    Autonomous entities Observe through sensors Acts upon environment using actuators
  10. 10. Intelligent Agents    Autonomous entities Observe through sensors Acts upon environment using actuators
  11. 11. Emotions    Change parameters based on events Aggressive, Defensive, Panic, Revenge... Class Strategies
  12. 12. Platformer  Player vs World  Run, jump and shoot
  13. 13. Trial and Error     Limited number of options Calculate best from many tries Pick best or retry (TAS) AI can try without actually acting
  14. 14. Real Time Strategy / Sports   Player (human/AI) controls a team Each unit is also an agent
  15. 15. Self-Organizing Systems   Ants leave pheromones, path to food Equivalent to a shortcut
  16. 16. Shooter   Individual agents Big difference monsters and soldiers
  17. 17. Machine Learning   Machine Learning to study patterns Might classify player strategy
  18. 18. Neural Networks  Predict player behavior  Guard favorite spots  Ambush player
  19. 19. Fighting Game  Player vs Player  Remember combo's  Timing of buttons
  20. 20. Imperfection and Randomness     AI can input combo's perfectly AI knows everything about the world Add mistakes on purpose Add randomness in any function
  21. 21. Dungeon Master    AI controls entire game world Weather, characters, animals.. Even flow of the game
  22. 22. Conclusion    No single AI solution Define entities independent of human/AI Find the right strategy (or mix) for your game
  23. 23. Designing Intelligence Yoeri Staal @YoeriStaal info@staalmedia.nl

×