Coverage
2007-Lyon-BT

GDC Lyon Overall Impressions

Alex J. Champandard on December 4, 2007

GDC Lyon 2007 has just come to an end. It was the first of its kind in France, so it had little problems here and there, but overall I felt it was a success. With such “start-up” events, you get as much out of them as you put into them!

State of the Industry

Overall, I got a very positive feeling about the French and European games industry in general.

  • The European games industry seems in a very healthy shape — though not in a AAA northern American way. The game development scene in Lyon is surprisingly active too.

  • There are lots of jobs for AI programmers going — especially those with experience! If you’re willing to invest some time in those game AI skills, it could pay off.

  • France, compared to other countries of larger size, is witnessing a very healthy amount of technical innovation (from middleware to research, including in-house development). It’s probably due to the education system here…

  • Everyone cool knows about or reads this blog :-) Thanks for all your kind comments and welcome to new readers also!

My Behavior Trees Talk

Alex Champandard GDC Lyon 2007

Photo 1: Developers listening about behavior languages.

The talk went as well as I could have wanted it to go beforehand. Of course there’s still lots of room for improvement (especially in retrospect knowing the audience better), but that’s part of the fun! See the original photos on Flickr.

There was a row of seats left empty at the front, but people were standing in the back, so I guess about 60-70 people attended in total. It was after lunch, so nobody had the energy to walk out; nobody fell asleep and toppled off their chair either. I got four very good questions at the end also, so some people were paying attention!

Alex Champandard GDC Lyon 2007

Photo 2: Behavior trees are modular script interpreters.

Underlying AI Theme…

Mainly, the theme I took away from the conference was sacrificing AI design goals to ship the games! This was mentioned in both talks from the (rare) continental European AAA studios:

  • S.T.A.L.K.E.R. developers had to tone down their A-Life system to ship the game, and ended up with traditionally scripted areas with locked AI.

  • Crysis developers gave up on the idea of having top quality animation without foot-sliding. They just couldn’t connect it with the realities of responsive AI.

On the one hand it’s great that designers are thinking ambitiously about their AI designs, but at the same time it’s a shame that technology isn’t meeting those requirements.

Most of the other talks were focused on research. They were scheduled in a rather odd fashion in some cases, but I’ll try to cover them in more detail over the rest of the week!

GDC Lyon Post-Mortem

It’s the first year so there were little problems ranging from audio quality to missing sign-posts outside to help people find their way. But there was generally lots of good will going around, so things worked out.

Though December is a bit of an odd time for a conference, I have no doubt GDC Lyon will improve next year. In particular, I’d love to see more of the following:

  1. Roundtable, roundtable, roundtable! I think these are the most useful part of GDC. They get people talking and generally help establish contact with others of similar interests. My attempts to organize a Game AI roundtable somehow turned into a panel; the idea of a roundtable seems to have fallen out of favor since GDC Europe in 2005 where they also didn’t “support” that format. I think it’s a key feature of these “smaller” conferences, so it’s time to bring it back!

  2. More networking. I did admittedly miss out on a lot of Day 1 as it was pretty hectic for me, but there weren’t that many chances to meet other developers even on Day 2. Lunches were left up to the individuals, so people went their separate ways. Speaker lunches were shared with the Game Connection crowd, so it was easy to get lost! The big party at the end was rather loud and also integrated with the GC event, so it was hard to do any kind of talking. I had a few of the speakers I missed on the first day circled in my program, but I most of them I never managed to catch up with!

It was also hard/impossible to attend all the AI sessions; there were so many that clashes were unavoidable, and some rooms got filled out and closed up by the firemen responsible for security! I’ll have to find someone else to review those talks ;-)

Anyway, stay tuned for more detailed reports once I get back into my regular routine…

Discussion 1 Comments

Dmitriy Iassenev on December 5th, 2007

> S.T.A.L.K.E.R. developers had to tone down their A-Life system to ship the game, and ended up with traditionally scripted areas with locked AI. As I am A-Life system developer from S.T.A.L.K.E.R. game, I'd like to clarify this question. It is true, that the game was released with less A-Life influence to the gameplay as we initially wanted. It is true also that the A-Life is hard to debug and because of it the game was delayed . It is true, that the A-Life in the game is limited to the tasks it can produce for the player and after some time it can become boring to complete them. It is true, that some of the NPCs behaviours were not used in the release. It is true, that we do have some problems with the NPCs going through the levels (we need here RTS formation movement for nice view of group of stalkers going together). But it is not true, that S.T.A.L.K.E.R. ended up as the game with traditionally scripted areas with locked AI. I don't know who told you that. We even have no tools for traditional scripting (so we didn’t have the cut scenes in the game). There are several areas which needed to be locked for some time by the storyline. But after player passed them, they are open for A-Life system. In these areas, NPCs are restricted with their movement and decision making. But NPCs can still walk through the levels of the whole game (18 levels in the game) and you can watch them when they're doing this. They live in the locations, imitate life actions and migrate when needed. A-Life is a very rich concept and it is too big for realization in one game (especially FPS) and for not that big development studio as ours. That is why we didn't implement it by 100%, but we will try to extend our system in the prequel, which will be released soon, when the player will be able to see more clearly his/her influence on the game. I can apologize that A-Life demonstration in the game was not that much high impressive as it was probably expected, because it would prevent the developers from the concept (my congrats to Jeff Orkin who advertised GOAP very good).

If you'd like to add a comment or question on this page, simply log-in to the site. You can create an account from the sign-up page if necessary... It takes less than a minute!