[001]
CP44
Current Version
Changed line(s) 3 from:
n
A computer knows the location of its character(s) and the player(s) - it has access to the coordinates and simple math will tell it the angle to use, the distance, the time to take \
to:
A computer knows the location of its character(s) and the player(s) - it has access to the coordinates and simple math will tell it the angle to use, the distance, the time to take \\\"bullets\\\" to hit a target, etc. To make AI seem \\\"fair\\\", errors or inaccuracies have to be introduced to the behavior (giving correct code a chance to \\\"miss\\\" the target), or the computer is prohibited from making the calculations as quick as it really can (giving moving targets a chance).
Consider a tanks/artillery game... the computer knows how gravity is applied, what coordinates it is firing from, and where the player is. If wind is applied, it can be compensated for using the same numbers and units that the shooting will be done with. A human player is generally given a side-view screen, a concept that gravity will be applied (but not the specifics of the falling rate), a visual firing angle with limited increments (like whole degrees or 5 degrees, maybe lacking units/markers for the player to use), and normally variable firing power (maybe lacking units/markers, may also be hard to repeat the shot due to controls, like holding Space Bar - see Worms Armageddon). Games that are only mouse-controlled and don\\\'t provide friendly indicators and markers for power levels and angles are inherently less precise - and it\\\'s not the player\\\'s fault, these are design and interface issues. The fact that \\\"AI\\\" and the environment are mixed by a computer (even for third-party \\\"bot\\\" clients) means that a non-player-controlled character generally has access to the locations of other players even if they would not be visible to a human player.
Put simply: computers get to calculate (compute) their shots, humans (generally) get to approximate and try again if they\\\'re still alive.
About the only way to make AI fair is to restrict the computer to what humans get to interpret: an image on the screen and maybe some audio clues - deprive the AI of access to the game\\\'s coordinate system, gravity/mass/acceleration constants, and actual player position. This is obviously crippling for practical designs because at the time of writing, machine vision (shape/object recognition) is still an emerging science; and to have a reasonable computer-controlled opponent playing like a human, it would have to be able to interpret shapes, velocities, in-game sizes, possibly masses... all with a 2-dimensional pixel array just like humans do.
Consider a tanks/artillery game... the computer knows how gravity is applied, what coordinates it is firing from, and where the player is. If wind is applied, it can be compensated for using the same numbers and units that the shooting will be done with. A human player is generally given a side-view screen, a concept that gravity will be applied (but not the specifics of the falling rate), a visual firing angle with limited increments (like whole degrees or 5 degrees, maybe lacking units/markers for the player to use), and normally variable firing power (maybe lacking units/markers, may also be hard to repeat the shot due to controls, like holding Space Bar - see Worms Armageddon). Games that are only mouse-controlled and don\\\'t provide friendly indicators and markers for power levels and angles are inherently less precise - and it\\\'s not the player\\\'s fault, these are design and interface issues. The fact that \\\"AI\\\" and the environment are mixed by a computer (even for third-party \\\"bot\\\" clients) means that a non-player-controlled character generally has access to the locations of other players even if they would not be visible to a human player.
Put simply: computers get to calculate (compute) their shots, humans (generally) get to approximate and try again if they\\\'re still alive.
About the only way to make AI fair is to restrict the computer to what humans get to interpret: an image on the screen and maybe some audio clues - deprive the AI of access to the game\\\'s coordinate system, gravity/mass/acceleration constants, and actual player position. This is obviously crippling for practical designs because at the time of writing, machine vision (shape/object recognition) is still an emerging science; and to have a reasonable computer-controlled opponent playing like a human, it would have to be able to interpret shapes, velocities, in-game sizes, possibly masses... all with a 2-dimensional pixel array just like humans do.
