A question about the Future of AI in GAMES

[Groogy]

Besides what Torakka wrote.

There is also a difference in goal of the game. In Go the end goal is always for both opponents to defeat the other while in a game like EU4 or CK2 it might not be. Teaching an AI always requires a heuristic that tells the AI "You are doing good" as a way of scoring each time it plays so it knows what works and doesn't work. How would you apply that there in our games? Would it be every time the AI does world conquest it has done good? So then there would be one singular AI with only the goal of conquest and no personality. Lately since CK2, EU4, HOI4 and Stellaris we've worked a ton more on giving different opponents different personalities depending on in-game information which I find is much more important than AI doing a world conquest.

Of course the AI is supposed to make sure you have fun and if you just roll over him then you're not having much fun right? So it's not an excuse for not improving the strategical capabilities of our AI which we continue to work on but it isn't the singular objective with creating our AI.

 

[Groogy]

That has nothing to do with an AI being able to learn, that's actually probably harder to achieve going that route. Traditional AI with scripts to guide the AI is way better to solving a task like that. Heck even the potential of having implemented machine learning and then have it confide to that you kinda lose the entire point of why you implemented it in the first place.

Either way it was a rhetorical question, the point is moot until everyone have a cluster of computers at home

edit: Machine learning is a huge topic with a lot of different nuances to it and various different kinds of implementations. You have to think of these as tools, not solutions because they aren't. For instance AlphaGo was not just a purely self learning machine. It was using a MonteCarlo min-max which means it will create several thousands of potential moves randomly and evaluate which one of these are the most optimal in the long run. The exact details on how it all ties together with what it had learned from previous games/moves I don't know.