Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Henny]

Given that 99% of players of strategy games like Civ4 will never play multiplayer, it's strange how companies kind of drop the ball on what constitutes the single most important element of single player games, AI. I'm not a game developer though, so maybe it's just really hard making good AI, them having to cheat still strikes me as really cheap, though.

[dennislp3]

Cutting corners for $$$ is likely the reason. AI programming is a pretty in depth specialized field of programming and I can see why companies choose it as the common corner to cut.

Also no one has really raised the bar on the "industry standards"  when it comes to AI...so until that happens...we get what we get

[Sergarr]

It's hard making good AI because there is no standard "solution packet" a la graphic engine. You have to basically create everything from (mostly) scratch any time you want to implement AI in the game. It would be much better if there was a pure AI engine which would, from the position of a non-AI designer, work like a black box, with standardized inputs and outputs.

[EnigmaticHat]

The thing about AI is, unless you want to try and be a genius and invent some crazy new system, it isn't enough to design a good AI.  You have to design a game around the AI, not the other way around.

Look at Galactic Civs.  Really well know for its great AI.  Want to know what it doesn't have?  Positional play.  Your military units are about as effective regardless of where they're fighting.  The fleet mechanic also ensures that no matter how many ships you mass in an area, each fight will only have a certain max number.

Guess what AIs are really bad at?

Take another example, diplomacy modifiers.  They're great for SP diplomacy because they make the AI's decision making transparent to the player.  But in MP diplomacy they don't make sense (because humans won't follow them) and they aren't necessary, because a human player will just tell you what they're thinking.  The issue is that if you put them in your game, there's this whole set of gameplay (which often costs ingame resources, for example GCs requires you to research diplomacy) that becomes completely irrelevant if you play MP.

[Levi]

Given that 99% of players of strategy games like Civ4 will never play multiplayer, it's strange how companies kind of drop the ball on what constitutes the single most important element of single player games, AI. I'm not a game developer though, so maybe it's just really hard making good AI, them having to cheat still strikes me as really cheap, though.

Yay, I'm the 1% that pretty much only plays it multiplayer.   

[notquitethere]

In theory I like hotseat multiplayer on turn based strategy games like Civ or Wesnoth... but it take soooooo long to play a game.

[Sergarr]

The thing about AI is, unless you want to try and be a genius and invent some crazy new system, it isn't enough to design a good AI.  You have to design a game around the AI, not the other way around.

Look at Galactic Civs.  Really well know for its great AI.  Want to know what it doesn't have?  Positional play.  Your military units are about as effective regardless of where they're fighting.  The fleet mechanic also ensures that no matter how many ships you mass in an area, each fight will only have a certain max number.

Guess what AIs are really bad at?

Take another example, diplomacy modifiers.  They're great for SP diplomacy because they make the AI's decision making transparent to the player.  But in MP diplomacy they don't make sense (because humans won't follow them) and they aren't necessary, because a human player will just tell you what they're thinking.  The issue is that if you put them in your game, there's this whole set of gameplay (which often costs ingame resources, for example GCs requires you to research diplomacy) that becomes completely irrelevant if you play MP.

The badness of AI in positional plays is actually mostly because people usually try to only consider the movement of a single unit at a time - from bottom to top. What they need is a reverse approach - from top to bottom. In other words, make AI work based on plans first, and situation second, not the other way around.

The AI should be already able to handle all those systems, really, because they're much less complex than, say, walking, or driving in a traffic lane, and modern cybernetic systems are already getting there.

P.S. I personally blame Chess for that, because it was the first game with an AI and in Chess you both have an extremely non-linear non-obvious dependencies between the optimal solution and the given position - so you really could only use a bottom to top approach, and also you can only move one piece at a time, which solidified in a type of AI which works good for Chess, but it's horrible for most other games.

[Bohandas]

Given that 99% of players of strategy games like Civ4 will never play multiplayer, it's strange how companies kind of drop the ball on what constitutes the single most important element of single player games, AI. I'm not a game developer though, so maybe it's just really hard making good AI, them having to cheat still strikes me as really cheap, though.

What does it count as if you play against yourself?

[Virtz]

All this talk of AI makes me wonder. Does anyone know of a game that'd utilize neural networks for its AI? Like not even to train it during the game, I mean make use of a pre-trained neutral network for its decision making. I'm kind of curious if there's any notable examples that'd suggest whether neural networks are a worthwhile time investment in more complex decision making.

[Sergarr]

All this talk of AI makes me wonder. Does anyone know of a game that'd utilize neural networks for its AI? Like not even to train it during the game, I mean make use of a pre-trained neutral network for its decision making. I'm kind of curious if there's any notable examples that'd suggest whether neural networks are a worthwhile time investment in more complex decision making.

There are a few games which utilize neural networks for their AI. It's usually pretty crappy, because there's no real standard process for making them learn. There are a few standard processes, but they can get stuck halfway through. And even with correcting for that it still learns at a speed of a dumb animal, and usually acts as one, too.

[alexandertnt]

AI which is cheating can sometimes be more fun, as overcoming an enemy with significant advantages over you can be quite rewarding.

The problem is when the AI is clearly brain-dead, which makes it much less satisfying to win against. It's particularly bad when the AI is clearly cheating to make up for the fact that its clearly brain-dead

The AI should be already able to handle all those systems, really, because they're much less complex than, say, walking, or driving in a traffic lane, and modern cybernetic systems are already getting there.

This depends on the game. Do not underestimate the complexity of games. For some game systems, a good and intelligent non-cheating AI can be a very, very complex thing to make.

Also keep in mind that driving and walking AI are much more advanced due to the countless millions of dollars spent on research by governments and industry. Naturally, game AI is not going to get anything near that, as a good CivV AI is much less useful than a car-driving AI.

but they can get stuck halfway through.

Are you referring to the neural network getting stuck in a local-minima during the learning phase? For pre-computed AI (where much more time can be spent converging on a solution), there are many techniques to avoid this.

[Sergarr]

The AI should be already able to handle all those systems, really, because they're much less complex than, say, walking, or driving in a traffic lane, and modern cybernetic systems are already getting there.

This depends on the game. Do not underestimate the complexity of games. For some game systems, a good and intelligent non-cheating AI can be a very, very complex thing to make.

Also keep in mind that driving and walking AI are much more advanced due to the countless millions of dollars spent on research by governments and industry. Naturally, game AI is not going to get anything near that, as a good CivV AI is much less useful than a car-driving AI.

I object to the bolded statement, because I have a counter example - Chess. It also had millions of dollars spent into researching AI into it, all for the sake of that idea of building a machine smarter than a human. Chess AI is completely useless outside of that purpose, yet there were specialized programs and even hardware built for it.

[alexandertnt]

I object to the bolded statement, because I have a counter example - Chess. It also had millions of dollars spent into researching AI into it, all for the sake of that idea of building a machine smarter than a human. Chess AI is completely useless outside of that purpose, yet there were specialized programs and even hardware built for it.

Ach, I was supposed to mention Chess as an exception, but forgot Chess is effectively "the" game of games, and is quite complex, so it's a popular thing for companies to show off their tech to the public.

[Sergarr]

I object to the bolded statement, because I have a counter example - Chess. It also had millions of dollars spent into researching AI into it, all for the sake of that idea of building a machine smarter than a human. Chess AI is completely useless outside of that purpose, yet there were specialized programs and even hardware built for it.

Ach, I was supposed to mention Chess as an exception, but forgot Chess is effectively "the" game of games, and is quite complex, so it's a popular thing for companies to show off their tech to the public.

It's also very unlike all other games in that every position is mostly unique, which make it almost impossible to reduce through strategy and approximation, which means that you've got to calculate everything that you can. Most other games can be reduced much higher by strategy and approximation, but can't really be "calculated" in the same way Chess can. Proof - the fact that normal humans can play those games and win without calculating all their physically possible moves beforehand.

[alexandertnt]

I'm not quite sure what you mean. Chess AI use strategy and approximation all the time, they have to, that's how they can make a move before the sun goes supernova.