Carmack Keynote at QuakeCon 2011

[obsidian]

This year, Carmack talks about the development of Rage and tying up the last bits of code, strengths and challenges on working on different platforms from PCs to consoles to mobile devices, code quality and analysis tools, difficult future for the modding community but focus on releasing source code. For Armadillo, he talks about going back to building (and blowing up) actual rockets at the expense of his pocketbook instead of working contracts.


Keynote:





Q&A Session:

[fKd]

hahaha poor guy at the end. console fail whale then carmack goes, "right, thats it. im out..." haha

[MKJ]

heh i totally forgot about the .plan files.

[AEon]

I really like listening to Carmacks talks. There are always interesting things he brings up, and unusual interpretations, that after some thought make sense, making you wonder why you did not see it that way till now. The technical engine stuff will mostly only tell engine devs something, but one usually gets want Carmack is saying.

Amazing that I would sit through 3 hours of "just" talk. No explosions, no action, no flashy cut scenes... we seem to be not totally media dumbed down yet. Clarity and intelligent straight answers seem to have their appeal, in all this posturing media babble without substance.

[obsidian]

What I enjoy from his keynotes in recent years is that he goes beyond being able to express what he knows and what he is working on, but also talks a lot about what he foresees in the near future and some of the tests he keeps doing every few years to feel out where we will be in the far future (voxels and raytracing). I think this shows the difference between just knowing how to code engines and knowing how to innovate for the future, a key component for future success. He has a particular vision that is unique in the industry.

[spookmineer]

Everytime he gives a speech, it's amazing that he can just stand there and talk for over an hour, without even holding a little piece of paper. It's structured and entertaining (I've seen movies that seemed to drag on and on and were shorter than this).
Did he seem more open about what's happening than before? He didn't hold anything back.

[obsidian]

That's because Rage is 99% done and there isn't much to leak that people don't already know. He was pretty tight-lipped about Doom 4 and other stuff.

I'm a little disappointed Rage won't ship with a Mac/Linux version, hopefully they'll add support in the future.

[fKd]

carmack and idsoftware. all class! these speeches are always fantastic.

also, nice to seeya AEon

[AEon]

Waves back at fKd... I see you are still at it mapping. Did "temp" ever make it into some final map BTW?

[Eraser]

It's structured and entertaining

While I am just as much in awe of his keynotes as you are, I do have to disagree with that it's structured. Carmack has the tendency to start a sentence, think of something else halfway through and jump on to that while never finishing what he originally was going to say.

Carmack IMO isn't very good at giving presentations. Recently he's not had it quite as bas as in the past, but he has the tendency to go on this technical mumbo jumbo talking spree that only 0.1% of the audience would understand. Did this in interviews as well. Dig up some of the .plan updates MKJ mentioned and you'll find a wealth of technical talk that goes way above your head.

But I digress. The thing is that the stuff he talks about interests me quite a bit and that keeps me focused through all three hours of him talking.

[AEon]

True... alas that is the problem when folks have such a good memory, they cannot be bothered to put up some bullet list of topics they would like to talk about. It would help the audience a lot, and probably Carmack as well if such a list was up there on the wall. Then again you could say this lets Carmack talk about whatever he wants... without the audience starting to whine that he did not talk about "promised" this or that.

I really liked the Q&A session, and other than the self-presentation of some folks who seem to use the opportunity to present themself, the questions were interesting enough and to the point.

Just yesterday I was wondering what I would have asked, and it occurred to me that I would probably have tried to say how important and great it is to have "compilable" bot support in Q3A, and if he had done any more research on multiplayer bot AI and if future games from id will see the return of bot support in the MP part of their games. Though if bot support goes the way of all the other things in gaming, adding bot support for a mapper will probably be a nightmare, just like the creation of "good" content is becoming more and more out of reach for a single, mildly talented mappers.

[Eraser]

I'd like him to talk a bit more about where he sees content creation go to. Creating assets for games becomes increasingly complex and time consuming. With Rage, it's basically impossible to create new mega textures. All this diminishes the ability for mod creators to create mods (or rather, total conversions). You'd really have to set up a team of dedicated modelers and level designers to be able to create something interesting with new content.

I wonder how, down the line in, say 10 years, this will impact in-house development as well. id Software took nearly 7 years to develop Rage. As more detail gets into games, it will take more time to create assets. How are they going to solve this problem? Because at some point, there is enough processing power to put almost limitless levels of detail into a game but to create that detail, it takes man hours. Is procedurally generated content the solution maybe?

[o'dium]

Its not impossible to create new mega textures. Its simple functionality. You may not get the source media used in Rage to use, but you can use your own, and its the same process.

Its certainly hard work, and its a huge step over previous offerings in complexity in content creation, but its not impossible.

[Eraser]

Maybe I should've been more subtle in that expression there. Of course it's not impossible, but the amount of work and specific knowledge it takes is huge. It's no longer possible to just doodle on something for a week or two and have an impressive looking level.

[obsidian]

Stamping the MT isn't hard. Carmack said just this in one of the videos above. He had some novice level artists start stamping away and adding detail to sections of the map. The difficulty is indeed the amount of source content that they created - a few terabytes worth - that they may or may not be able to distribute. These are basically source textures with diffuse and alpha, normal and specular maps used as stamps which are later compressed down to the single megatexture block of data.


If there's enough interest here for a Q&A session with Carmack or id Software in general, I can see if they would be willing to answer some questions.

[Infernis]

Please do, because I'm very interested in their views on content creation as well.

[spookmineer]

Regarding content creation, there is some hope that you can just scan objects in:

In the Q&A section, Carmack talked a bit about voxel engines.
Euclideon is a game engine development company, and their claims seem quite out there. I saw a youtube movie and wasn't sure.

Here is a 40 minute interview with the lead engineer, it also shows a real time demo (at about 22.20). Movie gets a little bit cheezy here and there.
Weird, he claims LOD is level of distance.

It's still early alpha and the claims sound revolutionary (it doesn't even use a GPU, it's still on CPU only) but it looks interesting to say the least, and it doesn't look like a scam.

[Kat]

/whispers @ Obsidian: you might want to split the topic at the mention of "Euclideon" before the topic gets out of hand... and it will discussing that shit *cough*

[fKd]

wait to see the animation systems. thats what most are questioning about unlimited detail.

rage editing, well.... sure your not gonna make anything at their scale. but surely you can make smaller areas/maps and megatextures? or even brush based terrain with stamps? gah, cant wait to look at idstudio! hurry up time!

but at the end of the day, as a mapper. until someone makes a better mp game. im not moving. and lets not forget ent plus

@AEon: temp01 is in the stack of to be completed.

[Silicone_Milk]

What I don't understand about "Unlimited Detail" is that the guy talking in the video puts a huge emphasis on the point of "polygonal models are flat looking (trees, etc..)".

However, he then says that they have a polygon converter so artists can continue working as they're used to working and simply convert over to "atoms".

The shape remains the same though yeah? So wouldn't we just have "flat looking" point cloud data models and end up doing a ton of work for zero gain?

(A cube is still a cube regardless of whether you're showing it as 12 triangles or 5 trillion "atoms")

[fKd]

only if you worked just from a poly-atom process, im guessing the point (heh) is that you can go so far beyond what is currently possible. if you are only using simple objects the whole thing is pointless tho.

[obsidian]

I think he means you can create sculpted high (millions) poly models and convert those to voxels.

[Kat]

only if you worked just from a poly-atom process, im guessing the point (heh) is that you can go so far beyond what is currently possible. if you are only using simple objects the whole thing is pointless tho.

It is. And without more detailed information all the fanboy comments and articles that have been popping up are disingenuous and don't actually mean anything in a broader context of tech that's no-where near being proved as useful beyond hyperbole and superlatives.

@ obsidian: they use point clouds. I actually get the distinct impression researching into this that they're lumping a lot of technology 'talk' into the same conversation and not clarifying details on purpose... "voxels" are all the rage, "point clouds" aren't.

[Eraser]

Notch (developer of minecraft) on euclideon:
http://notch.tumblr.com/post/8386977075/its-a-scam

[obsidian]

Yeah, that confirms my assumption that they were relying heavily on instancing. I'm not so sure I would go so far as calling it a scam. I'm sure it's an actual product of their work, just that it may not work exactly as well as advertised for practical game development - but they're not going to tell you that.