Artificial Intelligence  
Lecture 01 - Introduction  
Edirlei Soares de Lima  
<edirlei.slima@gmail.com>  
What is Artificial Intelligence?  
Artificial intelligence is about making computers able to  
perform the thinking tasks that humans and animals are  
capable of.  
o Computers are very good at:  
arithmetic, sorting, searching, play  
some board games better than  
humans, ...  
o Computers are not very good at:  
recognizing familiar faces, speaking  
our own language, deciding what to  
do next, being creative, ...  
What is Artificial Intelligence?  
AI researchers are motivated by:  
Philosophy: understanding the nature of thought and the nature of  
intelligence and building software to model how thinking might work.  
Psychology: understanding the mechanics of the human brain and  
mental processes.  
Engineering: building algorithms to perform human-like tasks.  
Academic AI vs Game AI:  
Academic AI: solve problems optimally, less emphasis on hardware or  
time limitations;  
Game AI: entertain player, have to work with limited time and  
hardware resources.  
Academic AI History  
Early Days (time before computers):  
Philosophical questions:  
What produces thought?  
Could we give life to an inanimate object?  
First programmable computers (1940s): war simulation, break enemy  
codes, …  
Symbolic Era (1950s 1980s):  
Symbolic systems: set of knowledge (symbols) + reasoning algorithm;  
Expert systems: large database of knowledge + expert rules;  
Trade-off: when solving a problem, the more knowledge you have, the  
less work you need to do in reasoning.  
Academic AI History  
Modern Era:  
Increasing frustration with symbolic approaches (scalability problem);  
Move towards natural computing (inspired by biology or other natural  
systems):  
Neural networks (first suggested in 1943);  
Genetic algorithms.  
Key ingredient: ability to handle uncertainty.  
Current research:  
Machine learning;  
Big data;  
Deep learning;  
Current AI Advancements  
Google & Uber Driverless Cars  
Personal Assistants  
Autonomous Robots  
Game AI History  
Pac-Man (1979):  
Very simple AI technique (finite state machine);  
Semi-random decisions;  
scatter_time >= 5 (sec)  
Chase  
Scatter  
chase_time >= 20 (sec)  
player_energy  
time >= 10  
sec)  
player_got  
_energy ==  
true  
_
(
Frightened  
Game AI History  
Goldeneye 007 (1997):  
Sense simulation system: characters could  
see their colleagues and would notice if  
they were killed;  
Still relying on finite state machines with a  
small number of well-defined states;  
Sense simulation was the topic of the  
moment:  
Metal Gear Solid (1998);  
Thief: The Dark Project (1998);  
Game AI History  
Warcraft (1994):  
One of the first times pathfinding  
was widely noticed in action;  
Warhammer: Dark Omen (1998):  
Robust formation motion;  
Emotional models of soldiers;  
Game AI History  
Creatures (1996) & Black and White (2001):  
The first time neural networks were used in  
a game;  
The neural network-based brain of each  
creature allowed them to learn what to do;  
Made AI the selling point of the game;  
Game AI Model  
Complexity Fallacy  
It is a common mistake to think that complex AI equals better  
character behavior.  
When simple things look good: Pac-Man  
Semi-randomly decisions at junctions;  
Player comments:  
To give the game some tension, some clever AI was programmed into the game.  
The ghosts would group up, attack the player, then disperse. Each ghost had its own  
AI.”  
“The four of them are programmed to set a trap, with Blinky leading the player into  
an ambush where the other three lie in wait.”  
Complexity Fallacy  
It is a common mistake to think that complex AI equals better  
character behavior.  
When complex things look bad: Black and White [2001]  
Neural Networks and Decision Trees allowed creatures to learn.  
When many people first play the game, they often end up  
inadvertently teaching the creature bad habits, and it ends up being  
unable to carry out even the most basic actions.  
Perception Window  
Most players will only come across some characters and  
enemies for a short time, which might not be enough for the  
player to understand the AI.  
Make sure that a character’s AI matches its purpose in the game and  
the attention it will get from the player.  
A change in behavior is far more noticeable than the behavior itself.  
Illusion of Intelligence  
“If it looks like a fish and smells like a fish, it’s probably a fish.”  
if the player believes an agent is intelligent, then it is intelligent.  
For game AI the nature of the human mind is not the key  
point.  
The AI characters must look right and demonstrate intelligent  
behavior.  
Sometimes, simple solutions are enough to create a good  
illusion of intelligence.  
Halo [2001] increasing the number of hit points required to kill  
enemies made testers thought the AI was very intelligent.  
Illusion of Intelligence  
Player’s perception of intelligence can be enhanced by  
providing visual and/or auditory clues about what the agent is  
“thinking”.  
Animation is an excellent way to create a good illusion of  
intelligence.  
The Sims [2000] although it uses a complex emotional model for  
characters, most part the characters’ behaviors is communicated with  
animations.  
Triggering animations at the right moment is the key point.  
Illusion of Intelligence  
The goal of game developers is to design agents that provide  
the illusion of intelligence, nothing more.  
Game developers rarely create great new algorithms and then  
ask themselves, “So what can I do with this?”  
Instead, they start with a design for a character and apply the most  
relevant tool to get the result.  
Be careful to never break the illusion of intelligence:  
Running into walls, getting stuck in corners, not reacting to obvious  
stimulus, seeing through walls, hearing a pin drop at 500 meters, …  
Most Common Techniques  
Pathfinding  
Steering behaviours  
Finite state machines  
Automated planning  
Behaviour trees  
Randomness  
Sensor systems  
Machine learning  
Most Common Techniques  
Pathfinding  
Steering behaviours  
Finite state machines  
Automated planning  
Behaviour trees  
Randomness  
Sensor systems  
Machine learning  
Further Reading  
Buckland, M. (2004). Programming Game AI by Example. Jones & Bartlett  
Learning. ISBN: 978-1-55622-078-4.  
Introduction  
Millington, I., Funge, J. (2009). Artificial Intelligence for Games (2nd ed.).  
CRC Press. ISBN: 978-0123747310.  
Chapter 1: Introduction  
Chapter 2: Game AI