Difference between revisions of "Cheating artificial intelligence"

From TheAlmightyGuru
Jump to: navigation, search
(Definition)
Line 8: Line 8:
 
For the purposes of this page, "cheating" means an AI which is programmed to break rules. This is different than an AI which is not fairly calibrated to the skill level of the player. For example, poker-playing AIs have been designed that can play at a super-human level guaranteeing victory over any human. Although such an AI is not evenly matched to a human opponent, it is not cheating as long as it follows the same rules as the human players. However, if the AI were allowed to know which cards are held in each opponent's hand, something the human players are not allowed to know, it would be cheating.
 
For the purposes of this page, "cheating" means an AI which is programmed to break rules. This is different than an AI which is not fairly calibrated to the skill level of the player. For example, poker-playing AIs have been designed that can play at a super-human level guaranteeing victory over any human. Although such an AI is not evenly matched to a human opponent, it is not cheating as long as it follows the same rules as the human players. However, if the AI were allowed to know which cards are held in each opponent's hand, something the human players are not allowed to know, it would be cheating.
  
Also, "cheating" in this context only applies to multi-player games where the player and AI are expected to be evenly matched. There are a lot of platform games where the player takes damage if they land on spikes, but the AI-controlled enemies do not. Although this may not be realistic or fair, this is not an example of a "cheating" AI in this context. It's also not "cheating" if the AI doesn't accurately simulate a human body with all its flaws in sensory perception, delayed reflexes, and cognitive impairments, but it is cheating if the AI isn't limited to the same user interface as the player. For example, in many [[realt-time strategy]] games, a human player can only give a unit a command if they first move the screen over it and click on it, but the AI is often allowed to give all units commands at any time.
+
Also, "cheating" in this context only applies to multi-player games where the player and AI are expected to be evenly matched. There are a lot of platform games where the player takes damage if they land on spikes, but the AI-controlled enemies do not. Although this may not be realistic or fair, this is not an example of a "cheating" AI in this context. It's also not "cheating" if the AI doesn't accurately simulate a human body with all its flaws in sensory perception, delayed reflexes, and cognitive impairments, but it is cheating if the AI isn't limited to the same user interface as the player. For example, in many [[real-time strategy]] games, a human player can only give a unit a command if they first move the screen over it and click on it, but the AI is often allowed to give all units commands at any time.
  
 
In this context, cheating AI is used most frequently in strategy games, racing games, and fighting games. Many online databases which include a list of "cheating" AIs often use poorly-defined terms including games where the AI is just really fast and accurate rather than actually breaking the rules.
 
In this context, cheating AI is used most frequently in strategy games, racing games, and fighting games. Many online databases which include a list of "cheating" AIs often use poorly-defined terms including games where the AI is just really fast and accurate rather than actually breaking the rules.

Revision as of 09:55, 1 October 2019

A jump kick against AI-controlled Guile results in a flash kick cheat.

Cheating artificial intelligence is gaming terminology for when a multi-player game uses artificial intelligence to control a player which doesn't have to abide by the same rules as human players. Many video games which have an AI opponent incorporate some aspects of cheating because it is much easier to program than a competent AI which follows the rules.

Because good AI truly is so difficult to program, I'm generally okay with games which use a cheating AI provided even a observant player can't tell if it is cheating. When it's clear the AI is cheating, the game stops being as enjoyable to me. For example, in the fighting game Street Fighter II: The World Warrior, there are special moves that the player can only use by holding a direction on the D-pad for two seconds. However, the AI is clearly able to use these same moves instantly. I find this especially infuriating since the AI already has an advantage over slow human reflexes and clumsy input. Cheating AI usually causes me to stop playing a game because I just can't find it fun anymore. It's like playing chess when your opponent can bring pieces back onto the board whenever they want.

Definition

For the purposes of this page, "cheating" means an AI which is programmed to break rules. This is different than an AI which is not fairly calibrated to the skill level of the player. For example, poker-playing AIs have been designed that can play at a super-human level guaranteeing victory over any human. Although such an AI is not evenly matched to a human opponent, it is not cheating as long as it follows the same rules as the human players. However, if the AI were allowed to know which cards are held in each opponent's hand, something the human players are not allowed to know, it would be cheating.

Also, "cheating" in this context only applies to multi-player games where the player and AI are expected to be evenly matched. There are a lot of platform games where the player takes damage if they land on spikes, but the AI-controlled enemies do not. Although this may not be realistic or fair, this is not an example of a "cheating" AI in this context. It's also not "cheating" if the AI doesn't accurately simulate a human body with all its flaws in sensory perception, delayed reflexes, and cognitive impairments, but it is cheating if the AI isn't limited to the same user interface as the player. For example, in many real-time strategy games, a human player can only give a unit a command if they first move the screen over it and click on it, but the AI is often allowed to give all units commands at any time.

In this context, cheating AI is used most frequently in strategy games, racing games, and fighting games. Many online databases which include a list of "cheating" AIs often use poorly-defined terms including games where the AI is just really fast and accurate rather than actually breaking the rules.

Examples

This is a list of games which are important to me that feature a cheating AI, for the full list, see the category.

Title Released Notes
Diddy Kong Racing 1997-10-21 Uses rubberband AI allowing computer-controller opponents to accelerate faster than and exceed the max speed of their vehicles compared to when when humans are driving them. They also don't hit traps when off-screen.
Mario Kart series 1992-08-27 The series makes blatant use of rubberband AI by allowing computer-controlled opponents to accelerate faster than and exceed the max speed of their vehicles compared to when when humans are driving them.
Mortal Kombat series 1992-10-08 Several problems: The AI can perform charged moves without needing to charge them. The AI can turn and respond to teleport attacks before the human player can. The AI can see Reptile even when he is invisible.
Street Fighter series 1990-12-16 Most early games in the series allow the AI to instantly perform special moves which human players cannot. Charged moves (like Guile's flash kick) and multi-tap moves (like Blanka's electricity) can be pulled off instantly.
Warcraft II: Tides of Darkness 1995-12-09 The AI is not affected by fog of war, so it can always sees the whole map. The AI can also control all units all the time, while the player must find, select, and command each unit.

Links

Link-Wikipedia.png  Link-TVTropes.png