GDC 2009 – Monday and Tuesday AI Summit


The GDC AI summit was pretty good. A wide range of topics covered – although as someone noted almost all of it for bipedal creatures, usually humans specifically – so it didn’t have as much on strategic AI or for other areas like space/flying – things with 3 dimensional movement.

A good highlight was a great demo by Damian Isla that showed an AI searching for a player, getting confused when the AI didn’t see the player where it last thought it was, then exploring further afield as necessary. The small amount of behaviour gave some pretty nice stuff – a problem being that showing that behaviour to the player is very difficult, and the technical aspects of dividing up spaces to search can get pretty complex. The use of emotions like that though is immensely fun.

I must admit I am more into the behavioural and design side of AI then the technical implementations – I have my notes up from the days, but the notes for the technical sessions might not be as good. These will be up shortly – they’ll take a little time to edit, and I’ll add them here and make a post (this weekend perhaps) when I’ve sorted them.

As for slides – the locations of them will vary. GDC is locking down their public access to slides, and the AI Guild is going to be member access only to “people who have shipped one game” and are an AI programmer. Therefore I’ll probably have a look around for some slides I want to re-read from the author’s own sites.

3 thoughts on “GDC 2009 – Monday and Tuesday AI Summit”

  1. To be fair, the pathfinding talk covered mostly strategy — and the panels on expressing emotion were actually not that human specific. Also, *all* of the stuff we discussed in the animation lecture would apply to other types of animals easily.

    Flying, generally speaking, is a relatively easy problem in comparison to highly constrained indoor environments.

    There are many topics we didn’t cover, but there were opportunities at the end for open questions…


  2. Well, as was brought up in the AI Roundtable, space movement is a different pathfinding problem, and it’s a shame we didn’t have anyone from Spore describing how they did movement with any number of limbs 🙂

    If you covered everything in 12 hours of lectures, I’d have been impressed. It’d leave nothing for next year 😉

  3. There’s a paper about the animation from spore, and the pathfinding is outdoors so it’s easier.

    Spaceship flight is actually much easier to build a navigation system for, unless you’re flying around arbitrary dungeons — but that’s 0.01% of the games.


Comments are closed.