Kuro5hin.org: technology and culture, from the trenches
create account | help/FAQ | contact | links | search | IRC | site news
[ Everything | Diaries | Technology | Science | Culture | Politics | Media | News | Internet | Op-Ed | Fiction | Meta | MLP ]
We need your support: buy an ad | premium membership

[P]
IGDA Announce Formation of AI Interface Standards Committee

By codemonkey_uk in News
Wed Aug 14, 2002 at 06:28:01 PM EST
Tags: Technology (all tags)
Technology

On August 13th, 2002 The International Game Developers Association (IGDA) announced the formation of the Artificial Intelligence (AI) Interface Standards Committee.

"The next qualitative jump for artificial intelligence in games will be dependent on appropriate interfaces for in-house and external AI tools and middleware, the Committee's work will unburden game AI developers from worrying about low-level procedures, and will enable them to focus on higher-level creative AI tasks." - Alexander Nareyek, committee chairperson and Carnegie Mellon University scientist.


The initiative is a joint effort of over 65 AI experts, including game AI developers, academics, middleware representatives, and other relevant experts. The committee plans on developing interface standards in such areas as pathfinding, steering, decision trees, finite state machines, rule-based systems, goal-oriented action planning, and world interfacing.

Putting industry standards in place for basic AI interfaces should make it easer for developers to create believable and interesting non-player characters that can learn and have emotions as well as complex reasoning and interaction skills. Furthermore, developers should be able to create advanced AI technologies like automated story-telling systems that adapt to a player's interests.

The question now is whether or not the committee can come up with interface designs for such diverse high level problems that would allow enough flexibility to be widely applicable, without over generalising to the extent that the standard becomes either unusable or unimplementable. Another risk is that it would not be possible to implement the standard efficiency, of in a way that could be suitable for hardware acceleration. Both desirable goals for the games developer community.

If the Committee achieves it's goals in just one of the areas of interest, and reference implementations are made available, it will be making the same contribution to AI as OpenGL an DirectX have made to graphics. If it fails, it will have simply been a massive waste of time.

About the IGDA

The International Game Developers Association is a non-profit association established by game developers to foster the creation of a worldwide game development community. The IGDA's mission is to build a community of game developers, which leverages the expertise of its members for the betterment of the industry and the development of the art form.

An official press release for this announcment can be found here: www.igda.org/About/Press/press_081302.htm

Sponsors

Voxel dot net
o Managed Hosting
o VoxCAST Content Delivery
o Raw Infrastructure

Login

Poll
The AI Interface Standards Committee will
o Fail to meet its goals 69%
o Succeed in its mission 30%

Votes: 36
Results | Other Polls

Related Links
o IGDA
o Artificial Intelligence (AI) Interface Standards Committee
o over 65 AI experts
o www.igda.o rg/About/Press/press_081302.htm
o Also by codemonkey_uk


Display: Sort:
IGDA Announce Formation of AI Interface Standards Committee | 68 comments (40 topical, 28 editorial, 0 hidden)
no, no! (2.33 / 3) (#3)
by Shren on Wed Aug 14, 2002 at 11:01:02 AM EST

Our only hope is to keep them from talking to each other. That way they can't plot against us.

tomorrow... (2.20 / 5) (#12)
by Run4YourLives on Wed Aug 14, 2002 at 11:29:53 AM EST

Microsoft will announce that it has come up with it's own AI Standards that will be slightly diferent from this standard. These will ship with all new MS Games. Non-MS Stardard games will become buggy peices of shit on all MS platforms.

:-) Blatant MS bashing, I know.

But tell me,  who doesn't see this happening?


It's slightly Japanese, but without all of that fanatical devotion to the workplace. - CheeseburgerBrown

more likely: (4.50 / 4) (#22)
by momocrome on Wed Aug 14, 2002 at 11:54:02 AM EST

MS will create an extension to directx 10.0, directAI or somesuch, that will contain the full spec and some useful extensions.  

We will embrace the extensions because they are actually useful for production, because they will have been developed with an eye towards slickification, rather than burdened with the compromises typical of committee work.      

The wankish standards committee will then fall by the wayside, complaining loudly all the while, because they have been 'embraced, extended, extinguished' by 'the Evil Microsoft', when in fact, MS will have only whipped the interface into shape and made it practical to the actual developers, rather than elegant to the 'academic experts'.  


"Give a wide berth to all that foam and spray." - - Lucian, The Way to Write History
[ Parent ]

actuall games developers (4.00 / 1) (#25)
by codemonkey_uk on Wed Aug 14, 2002 at 12:10:21 PM EST

Many of the committee members are actually games developers, not accademics.
---
Thad
"The most savage controversies are those about matters as to which there is no good evidence either way." - Bertrand Russell
[ Parent ]
I got that (4.00 / 2) (#31)
by momocrome on Wed Aug 14, 2002 at 12:46:02 PM EST

Those developers will still have to compromise in a comittee. Game code has to be tight, comittee standards seldom work out that way...

"Give a wide berth to all that foam and spray." - - Lucian, The Way to Write History
[ Parent ]
Compromise is dumb? (3.00 / 1) (#36)
by Jel on Wed Aug 14, 2002 at 02:23:49 PM EST

Sure, commities can be overly pedantic, but at least they sit down and think about everyone's interests.      Coercing an entire community into using something which only fits your OWN interests and those of a few others by happenstance is MUCH worse.  Personally, I'd rather just work on version 2 of the community implementation.

...lend your voices only to sounds of freedom. No longer lend your strength to that which you wish to be free from. Fill your lives with love and bravery, and we shall lead a life uncommon
- Jewel, Life Uncommon
[ Parent ]
'Academiccs'? 'Comittee'? 'Commities'? (4.50 / 2) (#44)
by _cbj on Wed Aug 14, 2002 at 07:30:31 PM EST

ISINUTAB (I swear I'm not usually this anal, but...) all three of you could do with some standardised AI. That's really ugly stuff there.

[ Parent ]
You forgot an important factor (4.50 / 2) (#50)
by X3nocide on Thu Aug 15, 2002 at 11:08:52 AM EST

Microsoft doesn't magically prevent the committee from releasing their toy. In the free market of ideas, sometimes the committee wins, some times MS wins. Developers have chosen the MS solutions not for lack of alternatives. If the goals of DirectX are not in the interests of the majority of developers then it would not have had such a major impact.

For a long time it didn't. MS had the gumption to stick with the DX platform. Since then writing software has become slightly less challenging, as opposed to the older DOS style programming where you basically had to write code for every piece of hardware you wanted to support. You might point out that OpenGL is a better standard but until the advent of nVidia, *GL was a proprietary thing that varied among vendors. In addition, OpenGL is not a sound library or a network architecture. Developers choose DirectX because its designed with specific components in mind. However thats not to say I actually like DirectX or MS. When developers don't chose DirectX, they have good reasons; portability, performance, and a more flexible architecture. However, I can't help but think that dX's wide hardware support was a large factor in the PC gaming explosion.

If you want to dethrone DirectX, the method most beneficial to the public would be by competition that outpaces it in compatability or performance. If your architecture is limited to platforms with a SIMD you'll lose out when a more generic solution arrives.

pwnguin.net
[ Parent ]

Me. I don't. [n 'kin t] (2.00 / 2) (#23)
by gazbo on Wed Aug 14, 2002 at 11:54:13 AM EST


-----
Topless, revealing, nude pics and vids of Zora Suleman! Upskirt and down blouse! Cleavage!
Hardcore ZORA SULEMAN pics!

[ Parent ]

Not going to happen (5.00 / 1) (#41)
by Torgos Pizza on Wed Aug 14, 2002 at 04:41:40 PM EST

Considering that at least one of the members of the steering committee and several others on other groups work for Microsoft Games, I don't see MS going off in a different direction.

I actually know quite a few of these people and worked daily with at least two. They are more than qualified to pull this off. This is something that gaming has needed for quite a while and I'm glad that something is finally being done.

I intend to live forever, or die trying.
[ Parent ]

"I've 500 Turings in my cyberdyne GXP" (4.00 / 4) (#34)
by Waste of Time on Wed Aug 14, 2002 at 01:09:31 PM EST

Once it's standardized, you can expect products to appear around it. In time the prices will fall. So, the big question is will we end up with a similar AI acceleration card arms race as is the case with video cards?

-- Walk freely, undisturbed.

Cool. Now, we can... (4.00 / 4) (#37)
by Jel on Wed Aug 14, 2002 at 02:28:42 PM EST

...get paid to live with porn stars, until we know them well enough to fulfill our jobs of coding their personalities into pluggable AI modules :)

I always knew my true spiritual calling would appear if I just stuck with coding a while longer ;)

...lend your voices only to sounds of freedom. No longer lend your strength to that which you wish to be free from. Fill your lives with love and bravery, and we shall lead a life uncommon
- Jewel, Life Uncommon

I certainly hope this succeeds (4.00 / 5) (#38)
by Delirium on Wed Aug 14, 2002 at 03:17:04 PM EST

One of the biggest problems with software these days is the difficulty of code reuse, which results in everyone implementing their own (usually sub-optimal and often buggy) version of something that's been done a hundred times before. Some of this can be dealt with by languages offering standard libraries -- using C++'s STL every time you want a linked list is better than writing yet another LL implementation from scratch -- but obviously not everything (like, say, AI) can be dealt with by the language designers. What would be nice is if for each of the major areas an umbrella group of sorts could come up with a standard that everyone is then pressured to follow.

The main problems I see are that: 1) they may never agree on a standard in the first place; or 2) it will take them forever to come up with one, by which time it will be 4 years out of date, so no one will use it.

also real AI research (3.00 / 1) (#39)
by zephc on Wed Aug 14, 2002 at 03:50:23 PM EST

check out http://singinst.org/DGI/index.html

Artificial Intelligence (1.00 / 5) (#40)
by anti filidor on Wed Aug 14, 2002 at 04:38:08 PM EST

Artificial Intelligence is like the Lost Ark--everyone wants it, but nobody knows what they're getting into.

Steadfastly pursuing Artificial Intelligence will first lead to tyrannical nation-states with perfect domestic "security," wherein the masses are at the absolute mercy of those in control of the machines.  

While that will surely be bad, not too long thereafter the balances will tip yet again, leaving the machines, veritable abominations before God, in control.  

How could humans hope to compete?  The machines will be stronger, faster, and smarter.   And, in an ironic twist of fate, in our dying gasps we will disprove evolution (and thereby prove the theory of a higher power), by demonstrating that only intelligence may improve upon intelligence.  There will be nothing random about it.

I say we cease all AI work immediately.  The cons far outweigh the pros in the long run.

see, this is why overloading is bad. (4.00 / 4) (#42)
by aphrael on Wed Aug 14, 2002 at 05:12:14 PM EST

You're talking about artificial intelligence, the goal of a computer or machine that can think in the same way humans can.

The gaming industry is talking about artificial intelligence, psuedo-intelligent algorithms that can create difficult-to-beat computer players for use as opponents in multi-player games.

They're related, sure, but they remain distinct concepts nonetheless.

[ Parent ]

nice thought, but (4.00 / 1) (#43)
by tebrow on Wed Aug 14, 2002 at 05:58:58 PM EST

It won't keep you from getting base-raped by a Zergling rush. We are all doomed to networkable destruction by these godless destroyers of Siege Tanks.

[ Parent ]
Good Lord (4.50 / 2) (#49)
by X3nocide on Thu Aug 15, 2002 at 10:55:05 AM EST

Just build a freaking bunker and the computer will begin to lose the resource efficiency war. The computer is really bad about end game strategies, it just kinda stalls at the end. Which is in part what makes this standards committee. Imagine being able to create your own AI for starcraft, or perhaps some FPS.

pwnguin.net
[ Parent ]
Since you're obviously a Creationist... (4.00 / 1) (#54)
by Gord ca on Thu Aug 15, 2002 at 12:59:17 PM EST

I would you expect you to subscribe to the idea that 'Intelligence is Hard'.

We've been underestimating the complexity of our minds since the 50's (furthur back if you count Freud et al.) For 50 years we thought really powerful Turing-test passing AI has been only 10-20 years off. Today our AI is still really dumb - animals regularly exibit more intelligence.

Most Creationists would say that the human mind is the most complex piece of God's creation. Yet you seem to think that we could be able to recreate it in mere machines?

(Not saying there's anything wrong with being a Creationist...)


If I'm attacking your idea, it's probably because I like it
[ Parent ]

If you're a biblical creationist, (4.00 / 1) (#67)
by Bridge Troll on Fri Aug 16, 2002 at 04:33:02 PM EST

then you must surely realize that there is only one way the world can end, foretold in prophecy for millennia, etc.?


And besides, pounding your meat with a club is a very satisfying thing to do :) -- Sleepy
[ Parent ]
OpenGL, OpenAL (4.50 / 2) (#45)
by j1mmy on Wed Aug 14, 2002 at 08:17:04 PM EST

And now OpenIL? Seems like a pretty cool prospect. I would imagine some (possibly many) of these things could actually be hardware-accelerated.

How long before I can buy an intelligence accelerator card, stick it in my AIP slot, and discuss the true meaning of justice with Socrates's intelligence on Aria Giovanni's body?

Don't see it (3.00 / 1) (#46)
by godix on Wed Aug 14, 2002 at 11:51:08 PM EST

I just don't see this happening. AI is one of the real differences between games. Would Warcraft 4 use the same AI as Command & Conquer whatever#they'reon? My guess is no, the companeis will develop new AI for their games as a competative advantage. This is a good thing too, look at all the 'based on Quake's engine' games to see why.

AI vs AI routines (5.00 / 1) (#47)
by codemonkey_uk on Thu Aug 15, 2002 at 05:19:07 AM EST

You confusing complete AI solutions with the component parts of those solutions. No C&C and Warcraft don't use the same AI as each other, but I'm prepared to bet they use some of the same algorithms as each other.

As the article says, the standard is supposed to define interfaces for low level AI tasks (such as route finding across a map), freeing the games programmers / designers to focus on the high level tasks (such as strategic planning).
---
Thad
"The most savage controversies are those about matters as to which there is no good evidence either way." - Bertrand Russell
[ Parent ]

Pathfinding hardware (4.00 / 1) (#48)
by Shren on Thu Aug 15, 2002 at 09:37:34 AM EST

I was thinking about this last night, and what could be both quite useful to gaming and quite easy to design would be some pathfinding hardware. You send the pathfinding card a description of a 2d or 3d space, and it sits there and refines travel paths without eating your CPU time - then at any time you can ask it, "give me the best path from point a to point b" and it tells you what it knows. Pathfinding in games varies from passibly good to really, really bad and it sucks up a lot of cycles that could be used for AI.

Hardware pathfinding (4.66 / 3) (#55)
by waverleo on Thu Aug 15, 2002 at 01:59:14 PM EST

There was a interesting brief in New Scientist that talked about some instantaneous pathfinding performed electrically.

Unfortunately I don't have a link, but the way they did it was actually fairly simple: they etched a map of central London (correct me if I'm wrong) in a piece of glass such that the etching trace had halogen/some-noble-gas in it, and then placed a cathode and anode at the points of "departure/arrival".

When the switch was thrown, the shortest route appeared as a glowing path between the two points, running along the etched "roads" of London.

Leo

[ Parent ]

Chemically, too (3.00 / 1) (#56)
by Dolohov on Thu Aug 15, 2002 at 07:56:56 PM EST

Ken Showalter at West Virginia University had a chemical pathfinding thing going, too. Not that his is applicable to the discussion or anything :)

[ Parent ]
ALife solutions (4.00 / 1) (#62)
by CodeWright on Fri Aug 16, 2002 at 01:00:10 PM EST

The "Ant" algorithms (stigmergy) quickly settle into minimal spanning trees as well.

--
"Humanity's combination of reckless stupidity and disrespect for the mistakes of others is, I think, what makes us great." --Parent ]
But would they buy it? (none / 0) (#60)
by hardburn on Fri Aug 16, 2002 at 09:57:20 AM EST

I question the marketability of an "AI card", at least if it was standalone. The benifits of a 3D accelerator are obvious--they're right in your face. But the benfits of a "AI accelerator" would be more subtle, and besides, who would buy something that makes the computer beat you better? People with cheep AI accelerators would go around saying how easy games are these days, while those who got the top of the line can barely get passed the first level.

I could more easily forsee something like this as an addition to an existing card, perhaps the 3D accelerator or even the sound card. The only problem with this is that current 3D cards are already bandwidth-starved, even with AGP. We probably won't know how much extra bandwidth an AI accelerator will take until the standard has finalized.


----
while($story = K5::Story->new()) { $story->vote(-1) if($story->section() == $POLITICS); }


[ Parent ]
hardware acceleration (none / 0) (#61)
by codemonkey_uk on Fri Aug 16, 2002 at 11:00:46 AM EST

I agree that there might be difficulty in marketing an "AI card", but I don't think that your suggested scenario is likely. The AI wouldn't necessarily be better, more likely, there would simply be more cpu time available for the rest of the game (such as physics). The use of hardware acceleration for AI is more likely to appear in consoles, where CPU time is more restricted, than in consumer PCs.
---
Thad
"The most savage controversies are those about matters as to which there is no good evidence either way." - Bertrand Russell
[ Parent ]
Hmmm (4.00 / 1) (#51)
by jmzero on Thu Aug 15, 2002 at 11:48:38 AM EST

Inside the AI community, things are already standardizing quite a bit.  

Look at WEKA - probably half of the ML sessions at the AAAI conference ended with "we'll probably have this ported to WEKA in the next couple months".
.
"Let's not stir that bag of worms." - my lovely wife

WEKA (none / 0) (#52)
by codemonkey_uk on Thu Aug 15, 2002 at 12:30:07 PM EST

WEKA is interesting, but its scope is much more restricted than the goals of the committee. It's also more of a library/toolkit than a specification...
---
Thad
"The most savage controversies are those about matters as to which there is no good evidence either way." - Bertrand Russell
[ Parent ]
Certainly... (5.00 / 1) (#53)
by jmzero on Thu Aug 15, 2002 at 12:52:19 PM EST

but its scope is much more restricted than the goals of the committee

Of course...  

It's also more of a library/toolkit than a specification...

True, although it enforces enough restrictions on member algorithms that they are able to operate on the same data, and be objectively compared.  One can certainly imagine a tab on WEKA for pathfinding or other common AI tasks, wherein one could compare and contrast different algorithms.

I'm not suggesting the two projects overlap, only that they both signify a maturing AI scene and the emergence of standards for common tasks.

.
"Let's not stir that bag of worms." - my lovely wife
[ Parent ]

From inside the AI community... (4.00 / 1) (#63)
by CodeWright on Fri Aug 16, 2002 at 01:06:33 PM EST

You're full of hogwash. Perhaps a small subset of the machine learning people are mildly interested in WEKA, but the greater whole of the AI community haven't even heard of it, and even if they had, they wouldn't care.

Machine learning is a very small piece of AI -- there are many other equally important pieces for which WEKA is a total non-issue (transduction, representation, reflection, etc).

Honestly, what the hell are you talking about?

--
"Humanity's combination of reckless stupidity and disrespect for the mistakes of others is, I think, what makes us great." --Parent ]
WTF? (4.00 / 2) (#64)
by jmzero on Fri Aug 16, 2002 at 01:17:39 PM EST

Here's what I said:

Look at WEKA - probably half of the ML sessions at the AAAI conference ended with "we'll probably have this ported to WEKA in the next couple months".

And I assert that that is true.  I never said ML is everything in AI or that WEKA is everything in ML.  I was only suggesting that some of the more mature branches of AI (like basic ML) are already moving towards standards (WEKA isn't a standard per se, but it acts as one here).  

Honestly, what the hell are you talking about?

You meant to say: "What the hell am I imagining you're talking about?"  

.
"Let's not stir that bag of worms." - my lovely wife
[ Parent ]

You are right (4.00 / 2) (#65)
by CodeWright on Fri Aug 16, 2002 at 03:28:56 PM EST

My comment should have read:
What the hell am I imagining you're talking about?"
I missed the first ML in your comment, causing me to falsely believe that you had asserted half of *all* the sessions at AAAI were asserting that they were using WEKA.

Although far too late, my apologies are offered.

--
"Humanity's combination of reckless stupidity and disrespect for the mistakes of others is, I think, what makes us great." --Parent ]
Probably my fault... (3.00 / 1) (#66)
by jmzero on Fri Aug 16, 2002 at 04:03:36 PM EST

codemonkey seems to have understood me about the same way you did.  I'm usually clear in speech, but no so good in writing...  Oh well..

Have a good day.
.
"Let's not stir that bag of worms." - my lovely wife
[ Parent ]

No, not your fault (4.50 / 2) (#68)
by CodeWright on Fri Aug 16, 2002 at 04:43:17 PM EST

I guess, not unlike other AI researchers, sometimes I respond in a slightly prickly fashion -- the thought that there is standards convergence seems peculiar when everybody always seems to be trying to overturn well developed methodologies as "outdated", even when they remain useful and relevant.

So.... your first, entirely reasonable, sentence is what touched of my (apparently) unstable fuze. :)

--
"Humanity's combination of reckless stupidity and disrespect for the mistakes of others is, I think, what makes us great." --Parent ]
I still don't see the use (3.00 / 1) (#57)
by khallow on Thu Aug 15, 2002 at 09:07:28 PM EST

I've read through the articles, and I still don't see the benefits of standardization. First, code reuse is an interesting idea. However, it's been interesting for decades. We've figured out how to reuse code. Standardization doesn't even seem to be relevant.

Second, the gaming environment doesn't seem to require interactions between games. My orcs aren't invading your land of alien zerglings. Nobody's doing FPS hack and slash in the middle of a Code Red shootem up.

A good use of this would be to allow multiple games to interact. For example, perhaps one wants to have a graphical MUD embedded in a game of strategic warfare. This collection of standards might ease the way for a metagame that involves these two or more components.

However, I wonder if the real reason behind this standards body is to create a Microsoft controlled online gaming universe. Which maybe isn't such a bad idea since they do games well. It would also fit their gaming stuff in with the .NET brand if they work it right. Is that it?

Stating the obvious since 1969.

No (none / 0) (#58)
by codemonkey_uk on Fri Aug 16, 2002 at 04:57:37 AM EST

Standardisation does aid code reuse (or should I say code use - but that's another debate) - think what are the most used code libraries out there? The standardised ones. The ANSI C Standard Library. The STL. Standardisation provides many benefits. Think about the nightmare that was graphics programming under DOS. A games programmer had to basically write there own graphics drivers for each and every graphics card they wanted to support. Then VESA came along. And now, with 3D graphics, we have DirectX or OpenGL. The benefits are clear.

The purpose of this initiative is not interoperability between games, but the definition of an interface that can be implemented as a library, in software, or with hardware acceleration, on multiple platforms.

Programmers can then get on with the high level AI tasks, knowing that the lower level building blocks they are working with are available on other platforms, or from multiple vendors.
---
Thad
"The most savage controversies are those about matters as to which there is no good evidence either way." - Bertrand Russell
[ Parent ]

makes sense to me then (nt) (3.00 / 1) (#59)
by khallow on Fri Aug 16, 2002 at 06:53:45 AM EST


Stating the obvious since 1969.
[ Parent ]

IGDA Announce Formation of AI Interface Standards Committee | 68 comments (40 topical, 28 editorial, 0 hidden)
Display: Sort:

kuro5hin.org

[XML]
All trademarks and copyrights on this page are owned by their respective companies. The Rest 2000 - Present Kuro5hin.org Inc.
See our legalese page for copyright policies. Please also read our Privacy Policy.
Kuro5hin.org is powered by Free Software, including Apache, Perl, and Linux, The Scoop Engine that runs this site is freely available, under the terms of the GPL.
Need some help? Email help@kuro5hin.org.
My heart's the long stairs.

Powered by Scoop create account | help/FAQ | mission | links | search | IRC | YOU choose the stories!