FEAR AI & physics, a decade without progress

[NikolaeVarius]

So what? That's good game design... recognizing what you have to work with and designing the game to fit that. I remember being impressed with the enemies in FEAR too, and I don't really give a shit what made it impressive. It worked well and I had fun.

It's asinine to call it "good" AI. Several devs have said that they have developed extremely complex AI for their own games that they simply couldn't use because they were too good. It's simply not fun to get flanked by AI all the time because they're outsmarting you. What happens is that you think the game is cheating instead of the game AI outsmarting you.

The premise of this entire thread is completely wrong.

 

[motsm]

It's asinine to call it "good" AI. Several devs have said that they have developed extremely complex AI for their own games that they simply couldn't use because they were too good. It's simply not fun to get flanked by AI all the time because they're outsmarting you. What happens is that you think the game is cheating instead of the game AI outsmarting you.

The premise of this entire thread is completely wrong.

I don't agree. Smarter AI just forces the player to think about their surroundings and tactical situation before engaging enemies. If developers have had to tone that down so brain dead gamers wouldn't rage quit, it's the fault of the gamer for thinking every encounter can be met with run and gun tactics. In other words, it's yet another situation of games being developed for the lowest common denominator, when they should be built to their full potential with an easy mode for those who can't handle it.

 

[Dankk]

IMO, AI doesn't have to be very smart in order to be good. If the AI isn't smart, but the game gives the player the illusion of the AI being smart, and that illusion upholds throughout the entire game, then it is a good game.

 

[Stringjam]

Play with humans if you want a challenge, if SP AI is boring you, time to give up on video games or instead of complaining about AI, get an education in programming and YOU go out and help create the next great AI.

I get complaining about a game cause it's got faults that we have already overcome in the past, but AI imo is either a long long way out or it's a pipe dream with expectations instilled on people by Hollywood movies that make it out to be way more than it is or even will be thanks to a lack of grounding to reality.

I don't get this kind of logic. "If you don't like something, divert your career path and become an expert in said field, and make it better."

Why don't we just shut down the whole forum then, or never discuss anything besides ponies shooting rainbows out of their asses?

Progress is made incrementally, built on previous successes and influence. We didn't go straight from lightbulbs to microprocessors. AI behaviors aren't any different. We don't have to jump from F.E.A.R. to Hollywood to make progress.

The point isn't "challenge." You can create an AI character that can respond to the player lock-step and shoot him between the eyes every time. The point is to create a SINGLE-PLAYER environment with interesting and balanced AI behavior. The requires creating A LOT of behaviors.

I just gave an example earlier of a developer who expanded on the behavior set with one game, and then the next game abandoned those behaviors. THAT is what I find puzzling.

Why not move forward? Keep building the behavior tree and branching out to more varied, interesting behaviors? There is no reason not to, other than resources. It takes a lot of time to build and debug, and if gamers don't seem to be interested in improvement, then they're not going to see the value.

I'm always happy to see AI discussion, and see no reason to throw out the "don't complain, play MP" card in an effort to halt the conversation.

 

[NikolaeVarius]

I don't agree. Smarter AI just forces the player to think about their surroundings and tactical situation before engaging enemies. If developers have had to tone that down so brain dead gamers wouldn't rage quit, it's the fault of the gamer for thinking every encounter can be met with run and gun tactics. In other words, it's yet another situation of games being developed for the lowest common denominator, when they should be built to their full potential with an easy mode for those who can't handle it.

Except, that was exactly what happened. Because the AI would utterly demolish any player in minutes in normal gameplay (Spam grenades and suppressive fire + flank = dead player), players would fall back into positions where they COULDN'T get flanked or suppressed permanently.

And then they would just wait for the AI to follow and then kill them. This happened constantly and every play tester reported it simply wasn't fun because the game would be just be inching forward and falling back constantly for hours.

 

[Topweasel]

Except, that was exactly what happened. Because the AI would utterly demolish any player in minutes in normal gameplay (Spam grenades and suppressive fire + flank = dead player), players would fall back into positions where they COULDN'T get flanked or suppressed permanently.

And then they would just wait for the AI to follow and then kill them. This happened constantly and every play tester reported it simply wasn't fun because the game would be just be inching forward and falling back constantly for hours.

The trick is to balance it give let an opponent flex it muscles without having it resort to having computer like targeting systems. It needs to replicate the actions of a computer. Not that of a human. You don't want a computer AI. You want a computer to fake trying to act like a human. Outside maybe two or three games (TLoU comes to mind) they don't even try that nowadays.

 

[motsm]

Except, that was exactly what happened. Because the AI would utterly demolish any player in minutes in normal gameplay (Spam grenades and suppressive fire + flank = dead player), players would fall back into positions where they COULDN'T get flanked or suppressed permanently.

And then they would just wait for the AI to follow and then kill them. This happened constantly and every play tester reported it simply wasn't fun because the game would be just be inching forward and falling back constantly for hours.

I still don't agree, but we never will, so no matter really. The first thing I would do in this situation is tone down the enemy accuracy, put a timer on certain AI actions, or even give the player more health or defenses. Removing AI abilities would be an absolute last resort however, unless they were seeing through walls or had super hearing.

 

[dust]

Except, that was exactly what happened. Because the AI would utterly demolish any player in minutes in normal gameplay (Spam grenades and suppressive fire + flank = dead player), players would fall back into positions where they COULDN'T get flanked or suppressed permanently.

And then they would just wait for the AI to follow and then kill them. This happened constantly and every play tester reported it simply wasn't fun because the game would be just be inching forward and falling back constantly for hours.

I still don't agree, but we never will, so no matter really. The first thing I would do in this situation is tone down the enemy accuracy, put a timer on certain AI actions, or even give the player more health or defenses. Removing AI abilities would be an absolute last resort however, unless they were seeing through walls or had super hearing.

AI auto balance can take care of that, and give better players a chance to play a challenging game.

 

[OVerLoRDI]

Fear is one of my favorite games of all time.

I played it through at least 5 times, and the last play through was on extreme difficult with no slo-mo. I rebound slo mo to some inaccessible key (like page up or whatever) so I couldn't knee jerk reaction use it. Super hard game without slo-mo, properly fast pace.

 

[tidy]

.

 

[WiseUp216]

gone

 

[YBS1]

I'm not going to knock FEAR, as on it's own I remember it as a pretty fine game. I have trouble looking to it as highly as some of you though as I remember FEAR as the game that killed off Monolith's (IMO) superior No One Lives Forever franchise. Now those games were fantastic, but from my understanding were not continued due to FEAR's greater commercial success. One thing to keep in mind however is we all do tend to paint a rosier picture of the past than it actually was at many times. Many times I've reinstalled an old favorite due to nostalgic yearning over better days gone past only to say "Ehh...screw this noise" 15 minutes in and go back to playing Call of Battlefield 8.

 

[Topweasel]

YBS I don't think that was the reason. If I remember correctly. WB bought Monolith right when FEAR was coming out. Monolith was already developing FEAR 2 before the launch. There was actually some issues with the franchise name (ip was Monoliths but the Name vivendi's) but they bought it off them.

What killed NOLF was a mix of a mid size studio having a new franchise and then getting purchased. WB wanted to use them as a launchpad into console games. FEAR 2 changed controls and actions to be more console friendly. AI was toned down. After that I want to say they started having them do DC games with Batman.

Never mind WB put them into even worse jobs. With Tie in jobs. Mordor is the first real new solo work since they let go of FEAR and condemned. Also I don't think NOLF did that well.

Ahh little more info. It basically became a rights mess after the WB purchase between Mono, Fox, and Vivendi.

 

[mizzou]

We've had numerous threads about this stuff. Any attempt to have emergent AI or advanced AI typically makes the chopping block before release because it has extremely unusual results. STALKER and Oblivion are two games that I know for sure had very unique goals with AI and ended up nerfing the entire thing.

Ultima Online even tried the same thing, but it messed the game up and ended up going completely away as well.

 

[Topweasel]

We've had numerous threads about this stuff. Any attempt to have emergent AI or advanced AI typically makes the chopping block before release because it has extremely unusual results. STALKER and Oblivion are two games that I know for sure had very unique goals with AI and ended up nerfing the entire thing.

Ultima Online even tried the same thing, but it messed the game up and ended up going completely away as well.

I works really well in games FEAR because is was a very Linear corridor shooter that they could engineer the combat areas to build up the great parts of the AI and limit the bad stuff that could happen. Its kind of funny though. Such great AI, but the slomo pretty much promised that most people wouldn't see it.

 

[tidy]

.

 

[werepossum]

It's asinine to call it "good" AI. Several devs have said that they have developed extremely complex AI for their own games that they simply couldn't use because they were too good. It's simply not fun to get flanked by AI all the time because they're outsmarting you. What happens is that you think the game is cheating instead of the game AI outsmarting you.

The premise of this entire thread is completely wrong.

I have no problem believing that several devs have SAID that they have developed extremely complex AI for their own games that they simply couldn't use because they were too good. I have a VERY hard time believing that is true. That implies that developers can accurately model human behavior, but we can't handle it so they fall back to AI which gets struck on building corners as it charges straight into our fire. Methinks if someone can develop such incredible AI, one could also develop AI which is enjoyably challenging without continuous immersion-busting stupidity. If nothing else, placing artificial limits on this supposed too-good AI would do the trick, giving us AI notably better than in F.E.A.R. but not so smart as to be unmanageable. This just seems like a "you're too stupid for me to not suck" defense of one's shortcomings as a developer.

I can see that in theory you have a point, but believing its practicality means believing that we eschew great AI like F.E.A.R. in favor of brain-dead Artificial Stupidity and we're just too stupid or too dishonest to admit it. Not buying it.

I don't get this kind of logic. "If you don't like something, divert your career path and become an expert in said field, and make it better."

Why don't we just shut down the whole forum then, or never discuss anything besides ponies shooting rainbows out of their asses?

Progress is made incrementally, built on previous successes and influence. We didn't go straight from lightbulbs to microprocessors. AI behaviors aren't any different. We don't have to jump from F.E.A.R. to Hollywood to make progress.

The point isn't "challenge." You can create an AI character that can respond to the player lock-step and shoot him between the eyes every time. The point is to create a SINGLE-PLAYER environment with interesting and balanced AI behavior. The requires creating A LOT of behaviors.

I just gave an example earlier of a developer who expanded on the behavior set with one game, and then the next game abandoned those behaviors. THAT is what I find puzzling.

Why not move forward? Keep building the behavior tree and branching out to more varied, interesting behaviors? There is no reason not to, other than resources. It takes a lot of time to build and debug, and if gamers don't seem to be interested in improvement, then they're not going to see the value.

I'm always happy to see AI discussion, and see no reason to throw out the "don't complain, play MP" card in an effort to halt the conversation.

Well said, and I agree 100%.

IMO, AI doesn't have to be very smart in order to be good. If the AI isn't smart, but the game gives the player the illusion of the AI being smart, and that illusion upholds throughout the entire game, then it is a good game.

That's an excellent point, and somewhat ties in with NikolaeVarius's point (which I'm not fully buying.) What we want isn't reality so much as a satisfyingly challenging subset of it. There's a reason people say war is hell, as opposed to war is fun and challenging. We want the illusion of war, but fun and challenging, in the same we want cop games without annoyingly defeating laws or racing games with enough realism to make us feel we are world class drivers without requiring us to be world class drivers.

 

[lupi]

have them all, haven't played any yet.

 

[TheSlamma]

I don't get this kind of logic. "If you don't like something, divert your career path and become an expert in said field, and make it better."

Why don't we just shut down the whole forum then, or never discuss anything besides ponies shooting rainbows out of their asses?

Henry Ford hated horses, he did something about it. Okay you are right I was off base with my logic, it makes no sense and probably never lead to any other discoveries or progress either.

And speaking of logic that makes no sense, you just quoted me saying I'm fine with discussions I just didn't agree with the expectations of AI, but you spin zone it and say we should shut down forums? Please, don't make this so easy.

 

[werepossum]

Henry Ford hated horses, he did something about it. Okay you are right I was off base with my logic, it makes no sense and probably never lead to any other discoveries or progress either.

And speaking of logic that makes no sense, you just quoted me saying I'm fine with discussions I just didn't agree with the expectations of AI, but you spin zone it and say we should shut down forums? Please, don't make this so easy.

What would be the point of these discussions? Every possible game flaw could be countered with "You don't like it, go produce something better".

Especially in the OP's case, he laid out an aspect where no one needs to go and produce something better. The exact same developer has already done so - and for some reason, abandoned it.

F.E.A.R. (without using the slomo) has excellent AI. Its two successors get progressively worse.

 

[BFG10K]

One thing to keep in mind however is we all do tend to paint a rosier picture of the past than it actually was at many times. Many times I've reinstalled an old favorite due to nostalgic yearning over better days gone past only to say "Ehh...screw this noise" 15 minutes in and go back to playing Call of Battlefield 8.

Good games still hold up years later, and I regularly replay older titles for this reason. Sometimes with newer hardware and enhancements such as source ports and community content, the experience is even better than the original.

I recently replayed Command & Conquer Renegade (the FPS), and I still found the experience solidly pleasant. The last time I played the game was ten years ago, in 2004.

Right now I'm playing through Oblivion for the first time and it holds up extremely well for an eight year old game. I was also surprised how tight the input is considering its console association

 

[tidy]

.

 

[nitromullet]

I assume most people reading this thread already have it, but The original FEAR + DLC is on sale at GMG right now for $2.49. http://www.greenmangaming.com/s/us/en/pc/games/action/fear/

 

[Veliko]

I found FEAR to be incredibly repetitive and certainly didn't notice anything special about the AI.

Other than it having a creepy atmosphere in places it was very average.

 

[Subyman]

The process and mathematics behind AI routines have been well established for decades. Its not a problem of creating better AI fundamentals, but a problem of deploying the routines in the way some people want them. Its also the need to balance the general gamer's expectations with a decent challenge. Creating a huge behavioral web that only people playing on the hardest difficulty would experience is a hard thing for companies to devote money to.

Would it be fun? A typical real life fire fight has guys dropping behind cover and having a stand off for hours in a police situation or, in the case of the military, there are usually multiple actors on both sides. The set up of one man versus thousands never happens in real life, so we aren't talking about real life behavior here. I think that is a fundamental problem. How do you keep the player alive while giving a large number of enemies human behavior?

There are a lot of difficulties and I'm sure it can be done better, but I can see why we don't have huge leaps in noticeable AI improvements.