I'm not sure Mike is serious about this proposal, although it sounded like it in the talk. He knows that miners/users will never accept this. bitcoin is unlikely to change in so major ways. And if you solve the mapping of identity to computer in new ways you can potentially have better systems than proof of work. there is no need for all nodes doing the same computation N times. going forward nodes is an increasing less useful concept. In the future compute power will not be local, but on some server. So the P2P model is really going away to some extent. bitcoin is now 5 years old, and since then the rise of cheap servers has been probably the most dramatic change. soon "cloud" will be even more ubiquitous, although there are of course many security issues.