Would YOU people believe me if I said neural expert systems are the future for gaming

K

K.I.L.E.R

Retarded moron
Veteran
Just doing some study for my Intelligent Systems exam and I've read a fair bit about "stuff".

Anywho I've come up with the idea that neural fuzzy expert systems are the most feasible form of AI for games.

Neural systems can learn, fuzzy systems are great at mimicking human stupidity and expert systems have awesome reasoning.
Why not put them together to make a human AI in Doom 4?

I'm asking this because I want to know if it's feasible short-mid term.
 
K.I.L.E.R said:
Just doing some study for my Intelligent Systems exam and I've read a fair bit about "stuff".

Anywho I've come up with the idea that neural fuzzy expert systems are the most feasible form of AI for games.

Neural systems can learn, fuzzy systems are great at mimicking human stupidity and expert systems have awesome reasoning.
Why not put them together to make a human AI in Doom 4?

I'm asking this because I want to know if it's feasible short-mid term.
Click to expand...

IME the less predictable AI behaviour gets (obviously if good AI were predictable it'd be a lot easier to write...) then the harder it gets to balance the gameplay and make a reasonable game out of it.

Also for every really smart thing you see an NPC do and think "wow, what great emergent behaviour, this AI is really cool", you see half a dozen other NPCs do something mind bogglingly stupid and without a nice simple set of rules to tweak it can be hard to make them stop.

It's also a PITA to debug/test because there are so many possible outcomes it's hard to get reasonable coverage.

So to try to answer your question... I think learning systems might be feasible, but undesireable unless they're quite restricted in what they can learn. Using fuzzy logic and expert systems are already quite common place, I'm pretty sure some implementations will also tweak themselves at runtime to learn player behaviour. So in some sense your suggestion is already reality. If the learning was much more complex and pervasive and the AI had to be trained during development, then it might just make life difficult without really improving anything for the player.
 
Adaptive next-gen neuro fuzzy expert emergent data-driven physics-oriented is the only thing that can save the game industry from its imminent doom, and only a true master like Derek Smart can realize their true potential.
 
I gather when you take a class on Decision Tree learning or alpha-beta pruning game trees, you'll tell us it's the future of gaming. :)
 
Nah. Those are backward innovations that take us nowhere.

DemoCoder said:
I gather when you take a class on Decision Tree learning or alpha-beta pruning game trees, you'll tell us it's the future of gaming. :)
Click to expand...
 
You must log in or register to reply here.
Back
Top