There aren't that many unique last names out there and looking for a paper by someone with a popular name might be a pain. For instance, according to statistics, the author named Y. Wang publishes more than 10 papers a day, and obviously it's not the same person :-)
Also, sometimes people change their names, use nicknames, or omit some of their initials, so introducing such unique ID's will be greatly appreciated.
The idea was around for quite a long time, with a few attempts of realization (one of those being the ResearcherID by Thomson Reuters). I hope it will work out this time.