The author of this paper, Pedro Domingos[1], just wrote a book on machine learning titled The Master Algorithm[2]. In the book, he talks at length about the various elements that may serve as (part of) the basis for a "master algorithm" - a generalized learning algorithm capable of learning anything. Rule induction is one of the things he talks about in the book, so at least one expert seems to think this stuff is still relevant.
Of course, it might just be that this is his pet "thing" and for all I know he could be totally wrong, but it struck me as interesting enough to start doing some reading on. And hence this post.