So this is basically writing a set of rules for a given type of design, then allowing the computer to run over them and just make random things, checking that they fit within a set of parameters, and presenting them as ideas?
It's a neat idea, because it also allows you to take two given designs you like, and use a random set of parameters from each to create "offspring" designs.
I had some old music software that would do this. Combine two presets to get a new one with a resemblance to the previous two, but also new in it's own way.
I see it as a way to start the design process though, not end up with a final concept, but I need to actually look at this a bit.
Edit: His company Genometri has a trial version of a solidworks plugin to do generative design, gonna take a look at it.
I see comments all over about how this will kill designer's jobs, and that's a viewpoint I don't get. Something similar is happening with Photoshop CS5. It has a new feature that will ""intelligently"" remove objects from the scene, and people are saying this will destroy photographer's jobs. Yet if you spend a few minutes with it you'll find it does a fairly poor job, leaving artifacts and obvious duplicates everywhere. It does speed up the job of removing obiects, since there's now a bit less work left for the photographer/graphics artist, and it seems this is no different.
Also, the blog goes over user-generated content a little bit. End-users are already able to self-design a variety of things, design a t-shirt or cup online, make your own website online... Anyone remember Geocities?