No excuse for poor ai/ai tech article

[dexters]

There are real issues with the AI and or bugs preventing parts of it from working properly. But the more I play the more confident I am that there's quit a bit working and wroking right.

Unless you caught the AI with their pants down or got very lucky and have turtle civs like Ghandi and the ottomans, AI presents quite a challenge.

 

[Doctor Phibes]

It's easy to make a challenging AI. All the game devs need to do is make it cheat. A cheating AI is always the best route for single player AI.

Except that that isn't an improved AI, which is what this thread is about. Plus the cheating approach is almost inevitable, as it stands, at an economic/empire-wide level, but I think most players would consider it unacceptable at a tactical level (for example, AI units healing twice as fast as the human player's) - unless of course that is a civ's unique ability, but those work at all levels.

 

[LDiCesare]

There's been work on robots (automated cars are just that) for years. I remember seeing automated cars driving on circuits with a few special landmarks 15 years ago. The car, robotics, electronics, military industries have all been trying to work on this for decades.
On the other hand, each game has its own rules, and there is a lot that would have to be reinvented for every game.

Pathfinding is more or less the only thing game ai does rather well, and even then you can see professional packages that are unable to provide a simple "find path to area" routine, only "find path to point", even though it's braindead simple to adjust the code to do that (the algorithm needs no change).
But when it comes to the point of handling conflict, even for pathfinding, the ai falls down. If you can block the ai path, should it compute another path or try to keep going? If it does predictably the same thing, it can be abused. For instance, I remember one game where I blocked a CtP2 enemy stack at some 20 squares distance by just moving a lone unit from time to time. If that hadn't worked, I could have ambushed it since the path it took was obvious.

Even the highest level ai depends on rules which are fickle and change a lot. Evaluating that something is good is pretty hard. Many humans are just unable to think correctly of the way they evaluate a situation, and express it. AI developpers have to be able to express it, abstract it well enough that when the game designer changes everything halfway through the coding, not everything must be started again from scratch, and they must do this with little background, little support from the rest of the industry, and with big constraints on computing time and development time.

To get a good idea of the state of ai in the game industry, you can look at this paper.
It's quite interesting, but you can see if you browse the site that much of the current games focus is on procedural animation, on brains that tell the animation engine what animation to play and nothing about strategic or tactical issues.

Even in military simulation, ai is still pretty limited. It's used to provide automation, not automatic enemies. It's very much like the car ai, namely it's supposed to apply a doctrine at a low level, but is unable to make decent decisions at a strategic, or even high tactical level.

 

[Kruelgor]

If done properly, a cheating AI is unnoticeable by the player. All the player sees is a competitive AI, a challenging and fun experience.

Cheating AI is most definitely the way to go for a game like CiV 5.

 

[Doctor Phibes]

To get a good idea of the state of ai in the game industry, you can look at this paper.
It's quite interesting, but you can see if you browse the site that much of the current games focus is on procedural animation, on brains that tell the animation engine what animation to play and nothing about strategic or tactical issues.

Really good link, thank you. I liked this summary of his, lots of food for thought:

* The part of academia that's working closely with industry is getting a much better idea of what commercial games require. The many more outlets for information are indeed proving themselves to be benefitial for the community as a whole, though there's still room for improvement in actually using the information that's available already!

* Industry still does not put as much focus put onto AI than say graphics or animation, and I'd say there's insufficient interest. In the case of academic research, this translates into much less funding for game AI projects, compared to the variety of animation work that seems to have less trouble finding funding from industry sources.

* The same goes for government funding as well, which points to a problem of notoriety of "game AI" accross the board. Multiple Ph.D. students and research groups (who are doing the more relevant work) are in a tricky situation. Either there aren't enough post-doctoral positions available in this area, or the funding is drying up due to the prestige of the area (as measured by journals).

* It seem that industry is not particularly open to disruptive or even innovative game AI techniques that affect the design of games anyway. So the occasional ideas that come from academia that are already applicable are not necessarily taken on board... I had trouble justifying this to a local journalist, but that unfortunately seems to be the case!

We can see the priority given to graphics in Civ5, but of course that's also partly because graphics issues are readily solvable. The disconnect with academia is indeed a problem, an odd one really, because computer scientists are often avid gamers and many of the early games came out of CompSci departments. But the most intriguing idea is that AI techniques may actually be 'disruptive' to existing game paradigms. That could explain a lot...

 

[SirTurtle]

I see most of this discussion is comparing Civilization AI and Chess AI in terms of the differences in combinatorial explosion, limited information, etc.

However I'd like to point out some of the challenges that a Civilization AI programmer faces that a Chess programmer does not.

1. Poorly defined game rules. The rules of Chess are completely standardized. Civ changes massively between each major version. During the course of development game rules are continually being tweaked. This means that even if the AI performs okay on a specific build, it may not work well at all on a later build.

2. Limited resources. Much of the Civ development team will be doing things like graphics, UI, etc. and not working on the AI.

3. Unclear agreement on what makes a good civ AI. An AI that is better at winning will not necessarily be more fun. Testers and customers are just as likely to complain about an AI that is "too hard" as one that is "too easy".


All that said, I'm confused as to why the game doesn't just give the player a combat handicap on higher levels. I think this would work well alongside the production and research bonuses the computer opponents currently get. E.g. on King the AI could get +5% in combat against human players, up to +33% on Diety. This makes sense to me: the AI is not as smart at building and tech order as a human, so it gets a bonus to compensate. The AI is not as good at tactical combat, so it should get a bonus compensate for this as well.

 

[Doctor Phibes]

3. Unclear agreement on what makes a good civ AI. An AI that is better at winning will not necessarily be more fun. Testers and customers are just as likely to complain about an AI that is "too hard" as one that is "too easy".

Well that is what levels are for. Assuming, and that is a big if, you had a decent AI, you should be able to dial it up or down. How you would boost or handicap it would depend on the nature of the code, of course. It's clear, in any case, that Civ5 has dropped to the "too easy" level, and this seems to the a consequence of 1upt. It's 1upt that is responsible for so many people evoking chess in this thread, wouldn't happen with stacks.

All that said, I'm confused as to why the game doesn't just give the player a combat handicap on higher levels. I think this would work well alongside the production and research bonuses the computer opponents currently get. E.g. on King the AI could get +5% in combat against human players, up to +33% on Diety. This makes sense to me: the AI is not as smart at building and tech order as a human, so it gets a bonus to compensate. The AI is not as good at tactical combat, so it should get a bonus compensate for this as well.

It's one way to go. But my intuition, as I said earlier, is that people take AI cheating with a pinch of salt when it's doing it quietly behind the scenes (getting hammer/gold/happiness bonuses, for example), but on a tactical level, with stronger or faster healing units, they'd be outraged. I could be wrong, it might be worth a try.

 

[Kruelgor]

Chess is to Civ5 as Tic Tac Toe is to Risk.

It's comical to relate the two.

 

[Doctor Phibes]

Chess is to Civ5 as Tic Tac Toe is to Risk.

It's comical to relate the two.

Not when discussing tactical AI. Recognition of impossibilities can inspire people to look for what is possible.

 

[drachasor]

2. Limited resources. Much of the Civ development team will be doing things like graphics, UI, etc. and not working on the AI.

This is the biggest problem, I think. Civ needs a real team working on AI. From what I read Civ 5 just had two people (and I am not sure that's all they did). You'd really want like 5+ people, imho, all working on designing and optimizing the AI so it runs as fast as possible.

Civ has always devoted way too few resources to AI development though. Civ 5 also suffers from not enough time being devoted too.

 

[SirTurtle]

It's one way to go. But my intuition, as I said earlier, is that people take AI cheating with a pinch of salt when it's doing it quietly behind the scenes (getting hammer/gold/happiness bonuses, for example), but on a tactical level, with stronger or faster healing units, they'd be outraged. I could be wrong, it might be worth a try.

You may be right of course, I can't speak for others or the general Civ playing community.

The better solution is of course to improve the tactical AI to the point where it can compete with a human. I'm just worried that Firaxis will never have the resources to do this.

 

[drachasor]

You may be right of course, I can't speak for others or the general Civ playing community.

The better solution is of course to improve the tactical AI to the point where it can compete with a human. I'm just worried that Firaxis will never have the resources to do this.

Making sure it isn't a complete and utter moron would be a good start.

 

[Doctor Phibes]

Making sure it isn't a complete and utter moron would be a good start.

Amen to that. And also one thing that should be more readily fixable, unlike the tactical AI, is its direction at a strategic level, including what passes for diplomacy. It would help if the AI stopped picking fights it can't win, for a start. It seems quite incapable of figuring out what is in its short-term interest, let alone long-term.

 

[Surgeon]

Urm... the google car can have as much processing power as it wants. It cost probably more to develop than all of Civ !