I think it is promising, even if it is a 'closed building'. The idea behind this technique should easy enough to re-implement with little to no more information. That there's precedent (a working demo), makes the pursuit that much more predictable and attainable.

To add to that, ID licences it's game engines for other developers, and as the paper suggests, is actually targeting different CPU architectures.

While not currently open source, I doubt this progress will stay within the realms of ID.