I think what we need are apprenticeships. I.e. a prominent research group hires kids after the remedial first two years of traditional college curriculum (or something like it plus a couple of years in an entry level job), and has them work on the actual problems at hand, filling the knowledge gaps as needed. It is no longer possible to know much of anything, so we should specialize and diversify at the same time, meaning that these students should be working at the seams between the fields: EE/CS, CS/Statistics, CS/Genetics, CS/Mechanical engineering etc. Further specialization should occur along the lines of whether the student would like to be a researcher, or have a more employable and practical set of skills. I never got much value out of lectures: few people can sustain rapt attention for hours a day. Most of what I know comes from the books. What I did get a lot of value from is working with people who know what they're doing, and from doing things myself, perhaps with their guidance and help. If this is heavily emphasized in this new model, I'm all for it.