Game Development Reference
In-Depth Information
chapter 5
Random and Probabilistic
Random systems are easy to understand. Consider the coin toss that starts
American football games, the winner of which gets to choose to kick off or to
receive the kickoff. The coin toss is not influenced by any consideration that one
particular outcome might be ''better'' or ''more entertaining'' or ''preferred.''
Probabilistic systems, on the other hand, consider the odds. In the card game
Blackjack, the dealer for a casino hits on 16 or less and stands on 17 or more. This
simple rule is based on a known long-term outcome that can be mathematically
Can That Be AI?
Both types of systems appear to conflict with our working definition of AI. What
is the intelligent part of random? How does a fixed rule deal with changing
conditions? Random decisions can simulate human behavior. Humans get
bored and distracted and are subject to the urge to try something different. So an
AI that is predictable will seem less intelligent than an AI that is not predictable.
At the same time, an AI that randomly picks from equally good choices is just as
good in the long term as an AI that picks the first of equally good choices it
evaluates. Our Minesweeper AI from Chapter 4, ''Rule-Based Systems,'' had a
very reliable figure of merit for the available choices, but it could just as easily
have picked among the best choices by random selection. A similar argument can
Search Nedrilad ::

Custom Search