Skip to main content

The Art House

Autodesk's Andrew Brammall on competition, the importance of AI and the Uncanny Valley

One of the best-known 3D design tools companies in the industry - not just videogames, but also architecture, automotive design and more - is Autodesk, and early last year it announced the acquisition of middleware company Kynogon.

Here Andrew Brammall, formerly of Kynogon and now the European sales manager for Autodesk's Games Technology Group, talks about the thinking behind that acquisition, as well as the importance of AI to games and what still needs to be done to improve it. How did the Autodesk acquisition of Kynogon come about?
Andrew Brammall

Well, the middleware space changed dramatically when Intel acquired Havoc about 18 months ago. Until then a lot of the well-established middleware companies had been running along on a fairly regular basis, and if they did end up getting acquired it was either by other middleware companies, or going back to the situation with Renderware - who was acquired by EA in order to get access to the technology.

But I think that with that acquisition it was quite an unusual situation where you had sizeable companies coming into the space, and I think that they realised the potential was there to use middleware technologies in a much wider range of applications - not just in games and console development, but other applications outside of the game industry.

So Kynogon started talking to Autodesk probably at the beginning of 2008 and it was the case that the idea behind it was that they were very interested in getting more involved in the overall game production pipeline - obviously Autodesk has great experience in producing digital content creations, but they wanted to get involved much more in the runtime space, also for use in other things, like design visualisation, and so on.

From Kynogon's perspective Autodesk was a perfect partner, because they're a pure software developer - they have no hardware associated with their own products as such, so it means that we can be truly multi-platform. Obviously we have ongoing relationships with the console manufacturers and in PC development, so any tools that we produce can easily be used across the whole development space.

The other thing that was interesting from Kynogon's perspective was that Autodesk realised it had no experience about how to develop and support middleware products. When we were acquired in May, Autodesk wasn't just interested in getting access to our technology, but also in the experience of the company as a whole - so everybody that was originally employed by Kynogon are basically Autodesk employees. From the point of view of the Games Technology Group within Autodesk, what does the company's cross-discipline, multi-industry reach, add to the proposition? Are there other benefits?
Andrew Brammall

I think the main thing we've found really is an enormous experience in 3D content creation, or animation. Kynogon always specialised in looking at artificial intelligence in games, so we've got a lot of experience in looking at how you can navigate characters through a 3D world, but we've not had any significant experience with animation per se.

So one of the things we're always been very interested in is trying to provide a character-centric solution for technology, so the idea that with current generation consoles and looking forward into the future, one of the things that people are really looking for is having truly believable and realistic characters in games.

We've always had this idea of what we'd describe as a trinity, so you've got a combination of physics, animation and AI all working together to provide a truly realistic solution for characters in games. We found with the ability to work with the other teams in Autodesk, they've got a tremendous experience on how characters move and look, and how you can apply that in a runtime space.

One of our long-term goals is to provide some sort of a framework for these three elements to work together, although I should say that the physics part of that isn't something we'd look to work directly on ourselves, because there are very well-established physics solutions out there, as well as clients using their own technology. So we're looking at ways to make it easier to slot in the physics as part of an overall character solution. Every developer has a different studio set up, different preferences on tools - who do you see as your direct competition at this point?
Andrew Brammall

Interesting question - I think to some extent it's difficult to try and identify a specific direct competitor. Where we're going to go with the Games Technology Group isn't completely formed, but the idea is that we can try to provide a more all-inclusive solution for the platforms.

At the moment we're sort of between companies that provide a complete engine, like Epic for example, and niche middleware products like Havoc. We're almost between those two. The Uncanny Valley - the term that describes the point at which visual quality becomes so good that keeping up with emotional investment becomes almost impossible - seems to be the new frontier for videogames. What's your take on it as a concept?
Andrew Brammall

My take on that is that I do agree with the basic idea, but I think that previous discussions have been more about an art-based solution, in terms of the physical representations of the characters - one of the things we would look at is that art isn't the only thing that contributes to a character, so you can have this fantastic-looking character, but if they do something stupid, you're not going to believe them as a character.

That's one of the things we're going after as part of this character solution - so you can avoid characters doing things they shouldn't, either not moving in the right way, as people are so tuned to looking at human movement, but also in terms of the decisions and paths that they're taking. One of the things that's going to disrupt anybody's immersion in a game is if one of the other characters around them does something spectacularly stupid. Do you think there's enough research being done into AI in the games industry as a whole?
Andrew Brammall

I feel that the research into AI is very much driven along the lines of trying to find the right results for gameplay as it stands at the moment. The key challenges for AI from the perspective of most developers would be things like 3D path-finding, avoiding objects in the game and avoiding each other, but also being aware of the surroundings that they're working in.

Where I think there's definite scope for improvement would be to look at things like the behaviours and reactions of the characters - I think to some extent Kynapse has reflected that, if you look at the way in which the AI is designed. There are three tasks that it has to carry out - perception, or being aware of what's going on, a decision phase, and the action phase.

Kynapse as a product is very good at looking at the perception phase, and the action phase, but we don't provide any solution for looking at the decision-making part, because that tends to be something specific to a developer's game.

So I agree that further research on that aspect is needed to make characters more believable, and also I'd say it's something we're guilty of, not providing a technical solution for that part. People like to talk of videogames becoming more mainstream, and now visual quality has hit a certain level is it the AI that's holding games back from being fundamentally more accessible to the rest of the population?
Andrew Brammall

I think so - I saw a great presentation by a tech guy from Microsoft, probably two or three years ago, and one of the things he described was this concept that current game-players have particular assumptions when playing games, because they understand the technology involved in producing a game. So if you get a non-player come in, they won't understand why you can't do certain things that they might be able to in the real world.

One of the examples was the idea of destructible terrain, which is obviously something that's changing at the moment, but back then you could hide a character behind a little wooden fence, drive a tank up and fire at the fence, but it would still be standing there because the technology couldn't cope with the concept of being able to destroy it.

One of the things we've been involved with is trying to provide a solution where you can have destructible environments. If you look at a game like Battlefield: Bad Company, one of the differences between that and previous editions of the game is that you can destroy buildings, fences, anything, and it's interesting that it's changed the way that people would play that particular game.

Andrew Brammall is European sales manager for the Games Technology Group at Autodesk. Interview by Phil Elliott.

Read this next

Related topics