> Comments as a long time gamer/newbie programmer trying his hand at modding/AI
It's a bit of a shame on the doom3 front, while not wholy unexpected that they'd have very basic AI, due to what doom is. Still Carmack is calling it how he sees it, it's just a pity that such a 'big name' is making tohose comments.
id have shown they've got an interesting and solid navigation system in the form of aas as shown by the quake 3 bots (even if they had to import it from Holland :). The aas being slightly unique since when it was designed for a quake 2 bot it was based around not having designer access to the map source, ie it had to work well straight off compiled bsps. So the compliler itself worked out reachabilities for legdes (walking off, jumping on), crouching, ladders, elevators, bobbing platforms, gap jumps, entering/exiting water.
Add to that that since the map is broken up onto many areas many techniques used for waypoint/navigation meshes as far as tactical pathing could be adapted.
Anyway other companies are working on doom3 tech so hopefully someone will actually do something with it.
As the Gamespy article (and various other accounts) don't quote him directly, I'm curious as to what exactly he said about "devoting resources to AI". It might not have been as dismissive as it sounds. Or maybe not.
Doom3 had really simple, deterministic AI that was a sort of pale, muddled imitation of the fantastic old-school patterns the monsters from Doom1/2 used. The brilliance of Doom1/2 was that each enemy was a building block for creating, together with level geometry, a unique and cool fight. Interestingly, it actually didn't have much at all to do with "scripting" (other than simple placement by designers)... the LDs didn't have much power to control events, so they just rigged up basic scenarios and let things happen.
Fast forward to Doom3 era (take for example the recent FEAR demo), and enemy encounters are usually very tightly scripted affairs, with designers often saying "go here" and "do this". Which is arguably even less about "AI" than Doom1/2 was. Carmack is right in saying that better algorithms won't help here, but he's wrong in assuming (if indeed he does) that that's going to be the dominant paradigm in AI for years to come.
Which kind of goes back to what we were talking about in a previous thread, where AI and design are going to be looking increasingly to each other for future advances. Ultimately I think this will prove very fruitful, but I'm not sure how much this will depend on new and more powerful technology or not.
It's long been speculated that Carmack's next engine will be his last. In a way, this kind of makes sense. Despite what his assertion that we're still not near the point of diminishing returns graphically, I think the increases in fidelity are starting to be lost on consumers even now as we hit the end of the current console generation. Only graphics programmers will care about the differences between Playstation 3 and 4.
What Carmack needs to do with his next engine is pass the torch to designers. His era, in which videogames rode Moore's Law out of the 1970s to greatness, is ending. In the future things will be more and more about the creative soul of the game, and on the technical end, working smarter rather than harder, creating new experiences via systems construction.
Posted by chmmr at August 15, 2005 01:55 PMTo be begrudgingly fair to Carmack, the summary of his talk does say this:
"...when applied to games AI is usually a matter of scripting. What game designers want is a way to act as the 'director,' telling enemy and friendly characters where to stand and what to do."
Maybe it's just time to get some new designers, John. :)
Posted by Borut at August 15, 2005 10:22 PMMy first post should have stuck to just: Further irony is seeing a bot class navigation system coupled with oh so simple AI.
Anyway it's fully worth tracking down the video of his speach. Running at an hour and a half I suggest throwing the audio on a cd and listening to it while commuting, as John doesn't do any particular acrobatics that are nesessary to see, winamp can throw it to wav via preferences-plug-ins-output-nullsoft disc writer.
John lead into this talking about how he knows how to use the graphics performance that will be available in the comming generations to make a better experience and how you'd use the multiple cores of upcomming cpus.
"it's not at all clear how you use incredible amounts of processing power to make a character do something that's going to make the gameplay experience better."
The quote about singleplayer AI that Borut posted is about here and leads on to.
"I keep coming back to examples from the really early days of the original Doom, where we would have characters that are doing this incredibally crude logic that would fit in like a page of c code or something. And characters are just kind of bobbing around doing stuff, and you get people that are playing the game that are beliving that they have devious plans, and they are sneaking up on you and they are lying in wait. this is all just people taking these minor, minor ques and incorporating them in their head into this vision of what they think is happening in the game. And the sad thing is you could write incredibly complex code that does have monsters sneaking up on you, hiding behind corners, and it's not at all clear that that makes the gameplay any better than some of these hapenstance that would happen as sort of emergant behavour of very simple things.
So, until you get into cases where you can think in games like the Sims, or perhaps massively multiplayer games where you really do want these sort of autonomous agent AIs running around doing things, but then that's not really a client problem, it's sort of a server problem where you've got large worlds there which again isn't where the multi-core consumer CPUs are really gonna be a big help on that."
I suggest this has a bit of a "640K ought to be enough for anybody" quality to it :) but then again I'm not someone to listen to.
Posted by Jarad Hansen at August 16, 2005 02:07 AM> ... And the sad thing is you could write
> incredibly complex code that does have monsters
> sneaking up on you, hiding behind corners, and
> it's not at all clear that that makes the
> gameplay any better
In a way, he's right -- if a creature's average lifespan is 10 seconds, it's very difficult to find anything you can do with AI that will help that creature add value to the gameplay within those 10 seconds.
As Borut put it, "Maybe it's just time to get some new designers, John. :)"
Posted by Paul T at August 16, 2005 08:26 AM"His era, in which videogames rode Moore's Law out of the 1970s to greatness, is ending"
we dont know what is in future .. maybe true 3D games where we can physically walk in simulated environments etc can even happen .. which may prove quite different challenges
Posted by shibinck at August 17, 2005 05:36 AMRegardless of what's in store for game graphics, there will come a point of diminishing returns, if we haven't reached it already.
As Nintendo President Satoru Iwata is fond of saying, "Someday game graphics won't look any better. What will we do then?"
Posted by Paul T at August 17, 2005 08:05 AMExactly. And most of the big challenges holding us back from doing "true 3D games where we can physically walk in simulated environments" are matters of user interface and human factors (and tech miniaturization, but that seems to steam ahead with time anyway).
If we could experience VR worlds today, Half-Life 2 level graphics would probably be quite adequate for creating visual immersion. It's the aural and especially tactile, input/output dimensions that need lots of research.
As Carmack said in his keynote, quoting an IBM engineer, "Graphics are basically done". Not to belittle the hard work of graphics programmers at all, but the road ahead of them is relatively straightforward.
Posted by chmmr at August 17, 2005 08:53 AMYeah, graphics are done and 64k is enough.
You could argue physics are done also since the road ahead there is pretty obvious, but I think you'd be wrong also. HL2 may be good enough, then again billboarded grass and trees still bring me out of the game world pretty quickly, as do low poly characters. If what you say is true, no one would be watching Xbox360 videos and saying "wow, that looks amazing, I want one."
Posted by SL at August 18, 2005 01:38 PMSL, could you clarify your comments?
Clearly no one is saying "64K is enough" (I guess this is a reference to the Bill Gates quote (of dubious authenticity) that "640K ought to be enough for anybody"), and I don't recall any low-poly characters or billboarded grass or trees in HL2 (though I may not be understanding your point entirely).
I do think there are a lot people out there who can't see much difference between videos of Xbox 360 games and videos of Xbox 1 games, even if the difference is obvious to those of us in the industry.
Posted by Paul T at August 18, 2005 03:29 PMPaul:
Yeah, missed the 0 in the 640. I used that statement as a blanket statement that tends to get made about how we've come as far as we need to and there's no point in moving forward. People always think we've made it as far as we need to go, then someone comes along and changes everything.
As for HL2, my point was just that claiming that what we have today is good enough forever just seems very naive to me.
We all know that graphics drive games sales as much as anything else, and while people may be happy with what is out there now, if they look back in 5 years they will wonder how they ever thought it was something good looking. You ever look back at GLQuake, easily the most amazing jump in graphics you've ever seen, and wonder how you could even make sense of those graphics?
Just because we find it compelling today does not mean we will find it compelling tomorrow. My mom, my girlfriend, and many people I know who are not hard core gamers by any stretch of the imagination are stunned by what they are seeing in the next generation of consoles. To say that the average consumer can't tell the difference just sounds, to my ears, to either be naive, or to be jealous of the fact that AI is never regarded with the same awe and wonder. I work in AI, so I understand that feeling, but to say "graphics are essentially done" is like saying that we'll never need hard drives bigger than 1Gig, or that no one will ever listen to real time streams on the internet (both things that I have read in the past by supposed industry experts).
Posted by SL at August 18, 2005 04:42 PM> I used that statement as a blanket statement
> that tends to get made about how we've come
> as far as we need to and there's no point in
> moving forward.
Fair enough, but I don't think anyone was actually saying that there's no point in moving graphics forward -- I think that's a misinterpretation of some of the comments made so far.
When the IBM programmer said that "graphics are basically done," I think he was saying that on the software side, researchers and developers have built so many optimal solutions to so many problems that there are very few problems left to be solved, and that's clearly correct. I think it's not too much of a stretch to say that great graphics from here on out will depend more and more on artists and animators and less and less on graphics programmers.
We clearly have this enormous momentum in constantly improving the visual quality of our games, and it's inevitable that that's going to continue, and no one can deny that it's a good thing.
Graphics are like money: no one -- including anyone on this blog -- will turn it down if you offer them more of it.
What we need to do is realize that there really is a point of diminishing returns, and if we want to move the industry forward, we need to focus on other areas -- most importantly, AI and game design.
> To say that the average consumer can't tell the
> difference just sounds, to my ears, to either be
> naive, or to be jealous of the fact that AI is
> never regarded with the same awe and wonder.
OK, when you say "jealous," I think you're just trying to insult us, and I'm not sure why.
I think "frustrated" would be a better way of putting it, since there's clearly an enormous gap between what's possible with AI and what most developers and publishers are trying to achieve (or perceive as possible).
Regarding my statements about whether consumers can tell the difference: I do think it's clear that *some* consumers are increasingle either unable to tell the difference or unwilling to pay a premium for what they see as a marginal graphical improvement.
Take a look at World of WarCraft and EverQuest 2.
When EverQuest 2 came out, not only was it relying on the massive EverQuest 1 player base and a near-total domination of the MMO market, it also had, in my opinion, a clearly superior level of graphical quality -- significantly higher polygon counts and (according to my understanding, though I may be wrong) a significantly more advanced renderer.
And yet ... according to my numbers (from www.mmogchart.com), World of WarCraft, starting from zero, has gone to over ***2 million users*** since it launched at the end of 2004. EverQuest 2, meanwhile, has a little over 250,000, despite the fact that EverQuest once ruled MMO-land with an iron fist.
So what happened?
If you were to judge by the quality of the graphics -- and the technical quality in particular -- you'd expect that EQ2 would be dominating WoW instead.
Posted by Paul T at August 18, 2005 05:22 PMMy intention was not at all to insult anyone, but I do think that frustration comes from the recognization of the fact that AI is not seen as being as important as graphics in games. Trust me, as an AI programmer, I share in that jealousy/frustration.
I think you are absolutely correct that good graphics alone will not sell a game. Looking at EQ2 I think it's obvious that the gameplay was not there and that you are correct in saying they were relying on their graphics alone. However, you look at two games of similar level of gameplay, where one has better graphics than the other, the one with better graphics will win every time. I think that's why you can look at RPG games imported from Japan which are still sprite based, and compare their sales to the Final Fantasy series, you will see that the FF games sell way better, and that series has been selling way more on their graphics than on their gameplay. Nippon Ichi makes RPG/strat games of a much higher caliber, but they rarely sell at the level of a final fantasy type.
In any case, I think that to say AI alone will sell games in the future is just untrue. Graphics will be the boat we sail on in, as we have always done. Good graphics bring people to the table, good gameplay/AI keeps them at the table.
Posted by SL at August 18, 2005 11:49 PMMy intention was not at all to insult anyone, but I do think that frustration comes from the recognization of the fact that AI is not seen as being as important as graphics in games. Trust me, as an AI programmer, I share in that jealousy/frustration.
I think you are absolutely correct that good graphics alone will not sell a game. Looking at EQ2 I think it's obvious that the gameplay was not there and that you are correct in saying they were relying on their graphics alone. However, you look at two games of similar level of gameplay, where one has better graphics than the other, the one with better graphics will win every time. I think that's why you can look at RPG games imported from Japan which are still sprite based, and compare their sales to the Final Fantasy series, you will see that the FF games sell way better, and that series has been selling way more on their graphics than on their gameplay. Nippon Ichi makes RPG/strat games of a much higher caliber, but they rarely sell at the level of a final fantasy type.
In any case, I think that to say AI alone will sell games in the future is just untrue. Graphics will be the boat we sail on in, as we have always done. Good graphics bring people to the table, good gameplay/AI keeps them at the table.
Posted by SL at August 18, 2005 11:49 PMBack to the original post for a minute. In reading what Carmack said in his speech, I think I detect a bit of truth in it. This may be me projecting a bit, and if you read my comments in the previous post you may agree, but I think what he was saying was to keep your AI simpler and to not work on ever increasingly more complex systems when the simpler systems with a nice GUI for designers to interface to it will have much greater impact than if you create large scale open ended adaptive AI systems.
Again, please feel free to point out that I'm wrong :)
I really don't understand how the two exhibits show irony.
Posted by aaron at August 21, 2005 06:50 PM"Those last two comments sure are demonstrating the state of the art of artificial intelligence!"
That last statement of mine sure is demonstrating why the original meaning of irony ("intentional, humorous double meaning") is the superior one!
Honestly... pedantically asserting that the original, narrower definition of "irony" is the only correct one is très passé -- and as much of an instance of comment spam as the two 'bot comments that -- *ironically* -- happened to follow it.
Posted by JayW at August 25, 2005 11:04 AM