Ph.D. allowed for some things: - You can deepen your studies and study the state of the art of a given area - You can work in some area with all the support needed (equipment, etc) - Networking However, what's happening is: - The latest journals are a click away - A lot of experiments can be done in a computer setting nowadays - Evolution in some areas do not depend anymore on experimentation, and/or the cost for it…
I think this is a very simplistic view that only holds for a very few areas of study. The idea that computer simulation is an adequate substitute for scientific experimentation in the majority of the sciences is puzzling to me. Not that models don't help, but there's still far too much we either don't know how to model, or don't have the computational power to model (or we have both, but it would be more expensive th…
Depends on the field. Even in natural sciences, there are several research areas not related directly with experimentation.
I would say that, for a lot of areas, we have the bases covered, but what's needed are better tools to understand it.
Protein folding is one area. The physics are ok, what we don't know is how to make it work for several atoms at once.
Data transmission is another. The real changes behind 4G, 802.11N, etc, is not a better radio transmitter only, but better channel estimation, error correction and use of diversity.
"I also don't understand why your fourth point means that specialization is not necessary"
It is necessary, but not sufficient. If you only focus on one area but don't know how to interface with the necessary surroundings you won't get anywhere.
" so I think it's at least premature to conclude that you can develop the same level of expertise as an autodidact."
The application of new discoveries and techniques is important as well, and several companies and individuals are doing important work in this area. (Go, Rust, several machine learning libraries, etc)