Transcript Evolution of Artificial Intelligence In Video Games

EVOLUTION OF ARTIFICIAL
INTELLIGENCE IN VIDEO GAMES
By Brian Bell
What is Artificial
Intelligence?
 The term Artificial Intelligence (AI) was coined by
John McCarthy in 1956
 As a Noun
 The Intelligence of Machines.
 As a Field
 “The science and engineering of making
intelligent machines.”
How does AI apply to games?
Overview
 Step through the History of Video Games
 Highlight certain AI techniques and trends
 Go over some of the pitfalls of modern AI
development in the Video Game Industry
Pong
 One of the First Video
Games
 Manufactured by Atari
 Released in 1972
 Designed for 1 or 2
players.
 Essentially just table
tennis in 2 dimensions.
The AI of Pong
 There is Logic to how the
Computer moves the
paddle.
 The computer predicts
where the ball with end
up based on a number of
factors then moves the
paddle to that predicted
location.
 To simulate Human error
the computer is designed
to skip its adjustment
every 8 frames.
Pac-Man
 Developed by Namco
 Distributed by Midway
(in US)
 Released: 1980
 Single player only
 Simple maze game
Ghosts
 While the Ghosts appear to move in a random order this
is not always the case. As each Ghost has its own
Algorithm to control its actions.
Chases Pac-Man
Try to get in front of Pac-Man
Moves at Random*
Super Mario Bros.
 Developed by
Nintendo
 Released in 1985
 Classic example of the
Platformer
Scripted AI
 Enemy units in the game
are designed to follow a
scripted pattern.
 Either move back and forth
in a given location or attack
a player if nearby
 Became a staple technique
for AI design.
Sim City
 Developed by Maxis
 Published by Electronic
Arts
 Released in 1989
 First in a series of
Simulation Games
which are often at the
forefront of AI
technology
A Living World
 The player is able to
manipulate certain
aspects of the city.
 The population of the
City acts somewhat
autonomously.
 Furthermore the game
will simulate random
disasters.
The Sims
 Developed by Maxis
 Published by Electronic
Arts
 Released in 2000
 A life simulation game
in which the player has
control over a semiautonomous people.
Complex characters
 The Sims characters
can be manipulated by
the player but will also
carry out day to day
tasks on their own.
 Characters also have
defined needs and
personality traits that
dictate how they act in
given situations.
Complex Object Interactions
 Characters are able to interact with an arrangement
of objects in the environment due to unique object
logic.
Black and White
 Developed by
Lionhead Studios
 Published by EA games
 Released in 2001
 Simulation game in
which you play God.
Black and White
 Similar to Sim City
 You had more control
over the environment
and the people in the
game
 You could also create
armies and wage war
AI Learning
 You could control and
train large creatures
 You could then punish
or reward the creatures
based on how they
complete tasks.
 Over time the
creatures will learn
how you wish tasks to
be done based on how
they are treated.
Combat Oriented Games
 Simulation games are
more advanced
 Combat Oriented
Games are more
Popular
 Evolution of AI in
combat games has
advanced quite a bit in
the past 10 years.
Halo: Combat Evolved
 Developed by Bungi
 Published by Microsoft
 Released in 2001
 First person shooter
that garnered fame for
competently brining
the genre to Consoles
Dynamic Enemy AI
 Enemies reacted in
dynamic ways.
 Different enemies
would react differently
in different situations.
 Enemies would work
together against the
player.
Behavior Trees
Tasks
Condition
Action
 Conditional Nodes
 Actions Represent actual
represent read only
functions.
 These functions are when
the AI unit observes some
information about its
environment
 Examples: Searching for
enemies or nearby cover,
determining orientation.
performed functions.
 These functions are
generally reflected in the
graphical representation.
 Examples: Firing on an
enemy, throwing a
grenade, reloading,
ducking behind cover.
Sequences and Selectors
Sequences
Selectors
?
1
1
2
2
3
3
 A sequence represents a
group of tasks that must be
carried out in order for the
overall task to complete
 A selector will ultimately only
carry out one task but
attempt to run each of its
child task until it finds the
correct task for the situation.
Behavior Tree
?
?
?
?
Behavior Trees:
Good and Bad
Benefits
Costs
 Allows AI units to act in
 Can be time consuming to
dynamic ways depending
on the situation
 Allows designers to group
different actions together
for different situations
 Allows for enemy
differentiation
generate
 More complex actions
require more complex
trees
 Running through all the
possible actions can take a
lot of time.
Behavior Trees
Bottom Line
 Are a good way to provide dynamic actions to
individual AI Units
 Generally require a larger AI control structure
for proper implementation.
Left 4 Dead
 Developed and
published by Valve
 Released in 2008
 A co-op first person
shooter set in a zombie
apocalypse.
AI Director
 The game utilizes an AI
director that will change
the game based on how
well the player(s) are
doing.
 This can make the game
more or less difficult for
players with different
skill levels.
 This also changes the
events of the game
between different play
sessions.
The Elder Scrolls V: Skyrim
 Developed and
Published by Bethesda
Game Studios
 Released in 2011
 Open World RPG set in
a fantasy world
Unbalanced AI
 Non Playable characters (NPCs) and enemies
within the game world react dynamically with
the player and each other.
Unbalanced AI
 Enemies are not designed to deal with simple
obstacles
Problems With AI in Gaming
 There are a number of issues that prevent
Video Games from being the best area for the
evolution of AI
 Despite the how far AI in games has come
over time no game is perfect.
 There are a number of reasons as to why this
is and will always be the case.
Sequels
Release Schedule
 Time
 The game has to be
released at some point and
not everything that needs
to be addressed can be
addressed.
 Money
 Games are expensive to
make so items are
generally prioritized to put
the most effort into what
is considered most
important.
Is it fun?
 Games are about
Entertainment
 While the AI in a game
might be very
advanced it might not
necessarily be
entertaining.
Are Video Games good for AI
Development?
The Good
The Bad
 AI is essential to Games
 Time and Money are an
 Games provide a simulated
issue
 Innovation might not be
that important
 Entertainment is more
important then quality
world where programmers
can experiment
 The Game Industry is the
largest area where AI is
being applied.
References
 Champandard, A. (2008, December 28). Behavior Trees for Next-




Gen AI. Retrieved May 2011, from AiGameDev.com:
http://aigamedev.com/insider/presentation/behavior-trees/
Champandard, A. J. (2012, February 9). Trends and Highlights in
Game AI for 2011. Retrieved May 2012, from aigamedev.com:
http://aigamedev.com/insider/discussion/2011-trends-highlights/
History of Game AI. (n.d.). Retrieved May 2012, from AI Game
Programers Guild: http://gameai.com
Montfort, N., & Bogost, I. (2009). Racing the Beam: The Atari
Video Computer System. The MIT Press.
Wexler, J. (2002). Artificial Intelligence in Games: A look at the
smarts behind Lionhead Studio’s “Black and White” and where it
can and will go in the future. Rochester, NY: University of
Rochester.