This depiction of game auteurs basking in the glow of game mega-auteur Will Wright (who writes the games I've always wanted to write) is a little odd to me. It seems to have an unhealthy obsession with Wright's technique of saving money by rendering graphics dynamically instead of paying someone to animate every possible sprite you might see in a game.
I'm no artist, so I've always accepted that I'd need to do this for any fancy-graphics games I might write. I've never had a budget for paying anyone for anything, and a "we need an artist" announcement is a near-sure sign of a moribund open source game ("Our ambitions have overshot our abilities! Who has some spare ability?"). So maybe it's just a different worldview.
The thing in that presentation that seems really interesting and novel to me is the tactic of creating in-game content by reusing the products of other users' normal play. Now that I write it down I see it's not totally novel: for instance, bones levels in roguelike games do the same thing. That's a lot more limited than what Wright is describing, though.
This type of hack, interpreting the normal behavior of users in a way that yields new information, is one of my favorites. In a game context it could cause privacy problems or be a Tar Pit From Hell, but you could structure games to avoid those problems; game state need not be a general-purpose communication mechanism.