David Sirlin on Chris Hecker's GDC 2008 talk
I’ve only seen Chris Hecker a couple times and both times he seemed like he was using some sort of illegal stimulant. Apparently, he is just always like this. I took Chris’s lecture as some sort of comedy experience or “ride.” After 20 minutes of highly abstract stuff he said “From here on out, it’s going to get a lot more abstract.” He said this with a straight face and I literally laughed out loud. He also said such lines as “I don’t know what this has to do with my lecture, or with games at all, but it seemed related (that was about Amazon’s Mechanical Turk service). He also said “If you can invent something better than the triangle, then unlimited money awaits you.” One of the questions at the end was actually “What was your lecture about?” and I’m not even making that up.
What his lecture was about is that there are few really hard problems we’ve solved in games that we solved really well. There is a similar character to these solutions. I won’t go into the details, but let’s just say they are awesome solutions. He talked a fair amount about “the triangle” being the biggest one, meaning a triangular polygon with a texture map. People tried all sorts of competing things like NURBS and other ways to describe meshes and surfaces, but the triangle apparently is the current king.
What he points out about this is that there’s a bunch of STRUCTURE to a triangle…the xyz coordinates, the uv coordinates, the way it connects to other triangles, and that it can have a texture map. Then there’s also the idea of the STYLE you can put on a triangle, namely the cool looking texture map. So programming people can play with all that first stuff because the computer understands the STRUCTURE of these triangles. Art people who know nothing about programming can play with the STYLE and create awesome 3D worlds and characters. Great solution!
He even said the triangle solution has had the biggest impact of any technoloyg in the history of games. But what SHOULD have had the biggest impact is AI. Too bad it hasn’t.
Chris says that AI needs a STRUCTURE/STYLE solution. There needs to be some way that we can define a structure of how behaviors in AI work, then let non-programmers define the style of creating behaviors for particular characters. He means something deeper than just messing with stats on a spreadsheet, but not something that involves writing real code. Do you NEED code to describe AI? He says his first answer was yet, but now he thinks maybe not.
To sum it up in a catch phrase, he wants “The Photoshop of AI.” A program that non-programers could use to create AI. He thinks we are no where near doing this now, but that it is possible. He said we’re far enough away that we’re better off not even trying explicitly for this yet, but on just generally understanding AI better first, and once we do, it will become more clear how to create that “Photoshop of AI.”
(This is excerpted from Sirlin.net post on GDC 2008, Day 1. It’s a great post, but this part is the most interesting to me. Go and read it!)
6 months ago