Stinky_Toe, you have a few facts wrong.
- HL2 uses soft shadows, not Doom3.
- Doom3 uses plenty of shaders. It just doesn't use the new shader features added in DX 9.0 or OpenGL 2.0.
- Doom3 does not require a modern video card (DX9/OGL2), but it will take advantage of NVidia's most recent OpenGL extensions for improving the speed of calculating shadows. (ATI doesn't have these extensions because NVidia copyrighted them.)
Doom3 IS cutting edge.
Just because Valve was exploring how to improve 3D visuals in a different area than id was improving, doesn't make it more-cutting edge than the other.
Valve is exploring ways of using pixel shaders for improving the image quality of water, specular highlights (that high-pass color thingy they demo'd) and stuff like that. They are still using traditional Lighting models however. (Light Maps) They use a few bump maps for texture details, but not normal-mapping like id is doing.
JohnC is exploring ways of improving 3D visuals by tackleing a whole new lighting model. The shader tricks used in Doom3 are probably no more complicated than Quake3's were, so John did not need DX9's or OGL2's new shader options.
That doesn't mean that Doom3's technology is not cutting edge. This stuff that John is doing has never been done before. (in a game)
Sure, there have been games (Halo) that have used bump maps, projected light textures or shadow volumns, but they were only used to tweak the eye candy. A traditional lighting model was still being used under all that.
For Doom3, these technologies ARE the lighting model. No more pre-calculated light maps, no more gradient fill triangles for dyanamic lights, no more inconsitancy between models, objects and maps.
Everything is done at run-time.