Adaptive/learning AI

Well.. I was not talking about online play... But I think adaptive AI may be the wrong word.. But I am looking for an AI that analyzes a game and "noticed" that I attack often from the rear of his base.. the next time he places more units there.. Or he noticed that I build alot of smaller bases, so he adjusts his playstyle and units to match.. or he noticed I used alot of long range units during mid-play so he adapts..

I know the AI will never truly learn.. but it is possible to "hardcore" certain tactics to help indentify certain tactics and know how to deal with it.

you mean like no AI in the history of RTS gaming? :lol:

The only way for an AI to be perfect is when it cheats.. I like to see a non cheating AI for various reasons

An AI will 'never' be perfect at the game. Can a human even be perfect at the game?

The best we can hope for is some form of pattern learning in the AI, so that if people say, frequently use certain choke points of the map, or a certain type of unit to overcome the AI, or take a specific route on the map etc, that the AI can adapt to some of those situations.

This data could be gathered from servers (with server hosts consent), and stored by Ubernet to create periodic updates a higher tier of AI. Something that will, from a players perspective at least, become tougher every update.

Ofcourse, since the worlds are to be procedurally generated, the AI will need to be able to analyse terrain to understand where chokepoints are, where vantage points are, etc, and react to players taking control of them, instead of blithely sending hundreds of units down a narrow pass held by the enemy.

sure you could get an ai like that, just have the game secretly send you into a multiplayer match with another fellow skirmisher. both will probably have a really challenging AI!

:lol:

seriously tho, the best you can probably hope for is an ai that cheats, and just sends excess units to what it thinks is your flank

if i remember properly the initial supcommm AI had a motto, something like, "If a first you don't succeed, try again, and again, and again, our units superior fighting spirit will overpower the enemy and win us victory", it just kept trying the same tactic again and again and again, any improvements over that would be very welcome, but i'm not holding out for it to pass a Turing test

I bet most of you couldn't pass a turing test. Your expectations are so far from reality.

I also want to be able to have a conversation with the AI where we can discuss the philosophical ramifications of sentient human-AI interaction. I will try to explain the scent of fresh mown grass, while learning about thedisconcerting sensation of a faulty transistor.

Perhaps PA will be the medium through which we learn to coexist peacefully with our robot brothers through discovering the shared joy of blowing each other up in a virtual world.

Would be fun.

Maybe, but would you pass a Millon test?

I'm not sure how it's relevant but yes I contain tyrosine.

I just want an AI capable of love.


And loving.


Bow-chicka-wa-wa . . .