GMAN-GAMES: The next false messiah

The next false messiah

年2004月12日09

I was hoping this article, "2004 GameTech Report: Creating Believable Characters", was going to be about story and or character behavior and A.I. having just played Half Life 2 in which most of the characters are unbelieable because of bad writing and bad acting. But, it turns out that's not what it's about

The article is about ways to dynamically generate motion. The links on the page will show you some examples, one of the most impressive in the last year has been the tech used in Endorphin. Basically they use some A.I. to give the character a goal and through physics and muscle simulation the character animates. You can set how fat, heavy, thin, lanky, the character is. You can set how strong their muscles are etc and they will act. It's pretty amazing stuff. It's a great way to create motions that would be impossible to create from motion capture. It's also possibly a great way to save time for regular motions.

But.............this is not the future of games IMO. While if it saves money and time it might open a few doors (e.g., before you couldn't put 8 main characters in your game because you only had time and budget to animate 1 character. Now with tech like this you can do all 8 in the same amount of time), will it really change any games?

Think about it. Imagine you had this tech that generated motion for you. Now, imagine you took that tech and you put it in Mario 64 or Jak & Daxter. What could you now do that you couldn't do before with those characters? Mario already runs, walks, crawls, hops, jumps, sneaks, slides on butt, slides on stomach, climbs stairs, climbs ladders, climbs poles, walks on ropes, hangs from ropes, hovers with a jetpack, rides a skateboard, hangs from ledges, climbs up ledges, etc etc. He already does all this stuff without this tech.

People talk about it making the feet not slide for example. While as an artist and an animator that bugs me, as a gamer I couldn't care less. Is my game going to be 1% more fun because the feet don't slide? I think not.

My point is, in the end, all this tech provides is better or cheaper animation but it doesn't appear to provide any new game play and that is the issue. If it doesn't provide new game play it's NOT the future of gaming.

Pass it on

コメント:

I have to agree that this kind of technology won't make your game even 1% more fun, but with the increasing complexity required by todays games, it may be the case that in the near future we won't be able to make any games at all without this tools. So maybe theyre not fun but they may be necessary.

Shorter-term, you might find that if you use this kind of tool you can dedicate more resources to gameplay instead of having a whole team animating mario's left eyebrow, or something. So it can make your game more fun I guess.

My two obvious eurocents.

posted by rebutoDecember 10, 2004 at 12:27 [ e ]

it's all about realism. Blow somebody's leg off and they fall on their ass in an interesting way. Try mocaping all the different ways to kill someone LOL.

physics gives anything replay value. Carmack's been building nice engines for the past 10 years, but the next 10's engines are going to focus more on breaking things up, and for that you need physics.

posted by anonymousTroyDecember 12, 2004 at 7:45 [ e ]

There are already games where you can blow someone's leg off.

Physics have brought some new gameplay but I don't agree prettier auto generated animation is going to bring the same. The important part is figuring out how to make the character choose interesting things to do, not how to animate those things IMO.

posted by greggmanDecember 12, 2004 at 11:16 [ e ]

ah, but those 'interesting things' benefit from unscripted and emergent behaviors, something that physics can really help with.

It's not just about blowing off somebody's leg -- it's about getting a whole cascade of realistic and potentially entertaining effects -- reactions with the environment -- for free.

Imagine a football or baseball game where the players were making totally lifelike body movement decisions -- it would be a revolutionary level-up to gaming, and that day is coming someday.

posted by anonymousTroyDecember 14, 2004 at 6:59 [ e ]

maybe we are talking about different things

The hard part is getting the character to decide what to do, not to animate that he decides. If a game could decide that stuff today there are already ways to get the character to look okay. That's why I brought up the examples of Mario and Jak. There's very little they can't already do. They can go anywhere, they can climb anything, slide, jump, run, ride stuff. So what if they can't limp, they could very easily move slower if they got damaged or we could easily add animations for limping or walking with an arm hurt etc. I agree that these new techs might make some of the animation easier but the real important part, the part that makes the game more fun or give the player more freedom have nothing to do with animation.

posted by greggmanDecember 14, 2004 at 9:22 [ e ]

yeah, I'm thinking more of NPCs than the player character (my sphere of interest is FPS...)

I just find your position on this curious since to me sim is the total future of gaming for FPS, sports, even stuff like Pikmin. Alas I don't see many games outside of CGW and Login so perhaps I'm a bit behind the times.

posted by anonymousTroyDecember 15, 2004 at 0:34 [ e ]

NPCs

I am talking about NPCs. My point is we can use Mario or Jak (or a Pikman) as an NPC today. The problem is not that we are missing animation, the problem is we don't know how to get him to decide to react to all the things he could react to. That is the harder the much harder problem.

Let's take Pikmin since that's probably a good example. What would you have them do that they aren't doing now? Whatever it is, it could be programmed (1) without extra animation, (2) with pre-made animation or with (3) generated on the fly animation. That's not the hard part. The hard part is figuring out how to get them to do what you want them to do. Build a bridge, drive a car, make babies, cough, hug, climb trees, dig holes, ride monsters, make rope, whatever. Once you figure out what you want them to do and how you want them to do it then you can use (1), (2) or (3) to give those behaviors animation. That hard part is not (1), (2) or (3). The hard part is programming those behaviors. That's the problem that needs solving.

posted by greggmanDecember 15, 2004 at 10:05 [ e ]

While you have a lot more experience in this area than I, the three years of PC game coding I did 1997-1999 was heavily into dynamics (car, aircraft, and weapon) and it was a blast.

The difficulties I saw with other parts of the games we did was that the character anims just didn't look *interactive* and natural enough, we always had to limit their actions in the world to what we had pre-canned animation for. Even first-class FPS like HL2 don't have enough resources to fully anim their NPCs and so they are much more limited in their actions than the PC (eg can't climb ladders). Getting arbitrary characters to walk believably on arbitrary slopes is one basic application, but the ladder example is where I'm coming from.

Previously I mentioned 'cascades' of realistic effects -- and this ladder climbing is a good example. It is possible to pre-can ladder climbing animations for the NPC, sure. But then you've got to pre-can everything else that the NPCs will do on ladders, like fire weapons, fall off, etc. Without a more general kinematic solution to the problem, the task of integrating an NPC into the game becomes larger and larger.

I agree that AI is always the bigger challenge, but I do think devoting more time into the kinesthetics of human action will pay dividends on the AI side too. For sports games I think we need to 'train' the NPCs to perform real-world athletic actions in a AI-directed, goal-oriented, forward-kinematic sense.

dynamics may not solve the AI problem, but ten years from now we'll be having a lot more of it. It's kinda like speech synthesis vs. pre-recorded characters, getting the output up to human level is daunting but once we get there we'll be golden...

posted by anonymousTroyDecember 16, 2004 at 1:55 [ e ]

I see alot of potential. Suppose you want the enemy to have a different physical reaction to being shot. Depending on where he's shot, his body is affected differently. Shoot him in the arm,and he grabs his arm exactly over the bullet hole. Or if you shoot him somewhere in the leg, he grabs that part of the leg. Since there are so many places to shoot a person and still hit their leg,it wouldnt make sense to write a different animation script for every single spot on his leg. Although that doesnt add much to the GAME, it does add alot to the experience that you're playing against a smarter AI (which involves the player in the world more than otherwise).

I can already think of alot of different ways to use this technology to make NPCs seem smarter without really writing AI.

posted by GarretDecember 21, 2004 at 2:53 [ e ]

[ コメントを投稿 ]

[ 掲示板 ]

All images not copyright me copyright their respective companies. Everything else copyright me.