Game Development Reference
In-Depth Information
with your own eyes. All you can do is
observe how people behave
in particular contexts. And when
you see them behaving in
specific ways
in
specific situations
, (like trying to avoid being attacked,
and trying to retaliate against an aggressor), we call that kind of behavior
intelligent
. And that's all
intelligence is: a behavior pattern.
The word
artificial
in artificial intelligence serves no purpose, because when a preprogrammed
character in a video game (like our Enemy) responds in
intelligent-looking ways
, even in a virtual
world, it's
not
demonstrating
fake
or
artificial
intelligence, it's demonstrating
real
intelligence. Its
intelligence is fundamentally no different from human intelligence. After all, in
both
cases we identify
the intelligence
by appearances alone
, and that's all we ever have to base our belief on. So, for
these thinkers (I'll call them
Functionalists
), there's no point asking whether computers will
one day
be intelligent. They're intelligent
right now!
It's only common misunderstandings that prevent most
people accepting it.
This Functionalist view runs against common sense today. Most people think there's a fundamental
difference of
some kind
between a human and a computer regarding intelligence. We feel the
Functionalist is missing an important piece of the puzzle. We feel that a computer, no matter how
sophisticated, will never be “truly” intelligent because it lacks an important, conscious and inward
ingredient that humans have. Now, whatever the case may be, there's something very useful in the
Functionalist view for the game developer, and also something very hindering and troubling in the
common-sense view. For this reason, whatever you personally think on the matter (I change my mind
every time I think about it), I recommend a pragmatic approach: suspend your position temporarily
and see intelligence from the perspective of the Functionalist. Why?
Often, when developers start creating a game with AI, they approach the matter believing their
aim is to develop some kind of
super-intelligent
and
truly clever
enemy; one that calculates and
figures out what to do just like a clever human would, and in just the same way. But because the
common-sense view maintains an intuitive feeling of incompatibility between human and computer
intelligence, it lures us into thinking that AI is something very difficult; into feeling as though we're
“out of our depth.” However, by taking the Functionalist view instead, and by seeing AI with
those
eyes, then new and exciting possibilities emerge. For game AI, the artistic rule applies: if it
looks
right, then
it is
right. If the enemy
appears
intelligent when the gamer is looking, then that's
intelligence enough. It's the appearance that we recognize as intelligence, despite what philosophical
debaters may have to say about “true intelligence.” If it's good enough to fool the gamer and offer
a believable experience, then it's “good enough” for us, and we needn't trouble ourselves unduly in
creating something beyond.
Note
There are some exceptional games that “go further” with AI and explore new boundaries and ideas,
and approach the task from new perspectives. These are often experimental games, “serious” games, or
simulators. But for CMOD and most other games, the Functionalist view will be our friend when coding AI in
C# and Unity.