Jump to content
  • Advertisement


This topic is now archived and is closed to further replies.


"honest" AI and "cheating" AI?

This topic is 6041 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

I''ve read an article that wrote, "Game developers tend to cheat their way through making AI". Actually, what are "honest" AI and "cheating" AI? What kind of cheating that can be done in Game AI, perhaps in a strategy game like C&C?

Share this post

Link to post
Share on other sites
If the reference is to literal cheating, then C&C and its heirs are famous for giving the computer advantages in initial cash, building speed, unit strength, and so on in order to make the game tougher This can be set in opposition to, for example, generating a computer general capable of marshaling its forces in flanking maneuvers. Or, in the case of RA2, one capable of using its engineers to repair bridges - not that I''m bitter or anything.

Otherwise, I would imagine that game developers are guilty of "cheating" in the form of parameter tweaking, but no moreso than with most other AI implementations.


Share this post

Link to post
Share on other sites
A common method of "cheating AI" in RTS games like Command and Conquer is letting the computer know your position at all times, or having a much smaller fog of war limitation. Sometimes they also get faster building times. Another common technique is to let the computer see the whole map so it knows where all resources are.

Most of the time, such techniques are done to level up the playing field since most AI''s are not up to human caliber when it comes to shifting strategies or responding to unexpected events.

I don''t generally mind computer AI''s cheating since I am usually better than they are, and I''d be afraid to see how weak the AI was without cheating at all. However, I feel that honest AI''s are more fun to play against because they often times make the same types of mistakes you do, and every once in a while you can really catch them offguard and see your strategies working. For instance, you can build all catapaults and they are unprepared. With a cheating AI, the computer migiht see you doing this an automatically compensate with fast units that can get to your catapults quickly. Cheating AI typically causes you to play with more balanced forces because they are always adjusting to your strategies.


Share this post

Link to post
Share on other sites
Heh, you guys have misinterpreted the statement just a bit

Game developers tend to cheat their way through making AI

What this means is that a developer designs an AI that "looks" intelligent. You see, most of the AI's used in todays games are just basic FSMs or FuSMs, nothing more complex than that. But the actions that the AI's carry out in the game make them seem to be more intelligent than they really are. In this way the developers have "cheated" in making the AI because it really isn't all that complex or smart of a beast when you really break it down.

However you guys have a nice discussion going anyways - I just wanted to set it straight that you were a little off topic In fact I'm surprised it wasn't explained further as such in the article. Oh well. That's what happens when you make assumptions. You should never make assumptions when writing an article. Anyways carry on...


Drew Sikora
A.K.A. Gaiiden

ICQ #: 70449988
AOLIM: DarkPylat

Blade Edge Software
Staff Member, GDNet
Public Relations, Game Institute

3-time Contributing author, Game Design Methods , Charles River Media (coming GDC 2002)
Online column - Design Corner at Pixelate

NJ IGDA Chapter - NJ developers unite!! [Chapter Home | Chapter Forum]

Share this post

Link to post
Share on other sites
Often, there''s no choice when it comes to designing AI for a game. AI take only a small percentage of the computer''s resources in a game, because all the rest are, unfortunatly, going to graphics, sound, and world computations. This leaves very little space for the AI to play with, thus they have to practically script events to make the AI seem like it''s smarter than it really is. Really, a pity in modern games.
If they were to make the AI with neural networks and complex reactions, it would make for a very poor game, as any AI further than FSM takes massive amounts of computing. Until we''re out of the binary computer bit, I think we''re stuck with either having a poor AI or a poor world when it comes to games.

Share this post

Link to post
Share on other sites

Well, you''re partially wrong here. The fun thing about NN''s, in contrast to GA''s and Hopfield Network, is that it''s completely precalculated, if you know what I mean. The time it takes to calculate an output vector based on a new input vector with a NN doesn''t take long at all. The Quake Neuralbot seemed pretty quick...


Share this post

Link to post
Share on other sites
Is there really a practical difference between AI that *looks* intelligent and AI that *is* intelligent? If AI looks intelligent *all the time*, then you''d want to class that as intelligent, regardless of the underlying mechanisms.

The real problem with game AI is that most of them don''t even look intelligent. But when you''ve got 20 artists on a game and only 1-2 AI programmers it''s hardly surprising that it looks better than it plays.

Share this post

Link to post
Share on other sites
Guest Anonymous Poster
I''m sorry, but when the hell did they actually start having
dedicated ai programmers on projects, much less two of them.

Oh, and depending on how you look it, they''re all state
machines, some of them are more complex and can change how
they''re structured is all.

Share this post

Link to post
Share on other sites
Guest Anonymous Poster
Ah, that''s just it, you accidentally asked two questions.
However, both are found within the thread but I shall explain
The quote means that developers cheat when making ai.
Putting honest and cheating into double quotes makes a
reference to the ai itself.

The quote from the article alludes to what Gaiiden was pointing
out. Gaiiden is right about current ai not bein complex but
he is wrong about the intelligence or smartness of it. Again,
smartness really isn''t needed since MOST ENEMIES ARE JUST
HAVE A LIMITED AMOUNT ACTIONS. You would expect an enemy to
shoot at you if it get''s close enough and very simple FSM work
wonders on doing that.
It would be like putting the mind of a genius into a coma

Honest and cheating ai are exactly that. You must remember
that the ai is another player.
Of course you could say that it''s impossible for the ai to
cheat because the advantages given to the computer are a part
of the rules of the game universe.

If you have an iq above that of a rock, then you should be able
to figure out the definitions.
If you''re worried about advertising, then celebrate whatever
technique you use to make the ai so smart it doesn''t need to
Happy New Year!

Share this post

Link to post
Share on other sites

  • Advertisement

Important Information

By using GameDev.net, you agree to our community Guidelines, Terms of Use, and Privacy Policy.

We are the game development community.

Whether you are an indie, hobbyist, AAA developer, or just trying to learn, GameDev.net is the place for you to learn, share, and connect with the games industry. Learn more About Us or sign up!

Sign me up!