The problem is not any "intrinsic IQ edges". These are mythological things that do not affect research results. As Mark Twain said, it ain't the things that we don't know, but the things we know that just ain't so. Courses teach you things that just ain't so.
The time spent asking questions about how to learn XYZ is better spent asking questions about XYZ. The way to do this is to consider a specific physical problem, and solve it. If you can't solve it, read some papers, and then solve it.
Nothing about AdS/CFT is taught in graduate school. The reason is structural--- there are always many more people who are interested in strings than there are who are interested in spintronics. So the graduate courses are structurally designed to push people out of string theory and towards something practical. The timeline for learning any amount of string theory in school is therefore $\infty$-months.
To learn AdS/CFT, read the original papers. They are very accessible if you have the string theory background. It is not good enough to read the reviews, even the best ones, you should learn the brane constructions of the 90s, because these contain an enormous amount of black hole physics. BFSS matrix theory is essential too, because this is a simpler arena pre-AdS/CFT, which forshadowed it, and I personally find the physics of BFSS more monstrously counterintuitive, although it is just a version of AdS/CFT. It is somehow more difficult to see 10 dimensions emerging than 1 plus a sphere.
Learning field theory today is not hard--- you can do it relatively quickly from some Wilsonian reviews. Field theory is useful outside of high-energy physics, and so is sometimes taught properly. Learning perturbation theory can take 1 day, or it can take 10 years, depending on whether you learn the path integral first. Lean the path integral first. Ignore anything Dysonish that does Wick's theorem by normal ordering time-ordered operators, and if someone is doing this, they are incompetent, and you can ignore all their research papers too. Dyson's methods were a stopgap measure to get physicists in the 1950s who refused to learn path-integrals to learn Feynman diagrams.
The central barrier is learning string theory, and to do this, one must read the original papers on dual resonance models, especially Veneziano's nice review of the early 1970s, and Mandelstam's. These are not accessible nowadays, because they require familiarity with S-matrix theory and Regge theory, neither of which exist as disciplines anymore. This is a terrible shame, but you can fix that by reading up books and articles from the late 1960s on your own. There is no way to understand string theory without going through Regge theory, which is why all the good string theorists are old. They protect their monopoly on knowledge by hiding Regge theory and S-matrix theory from the younger folks, although I am sure they do not do so consciously, only structurally.
The original papers on N=8 SUGRA used computer algebra to find the Lagrangian. The supergravity Lagrangian is so complicated, it is hard to use, certainly you can't use it for perturbative calculations because they are just too complicated in any formalism. The modern methods for calculating N=8 SUGRA amplitudes work them out using analyticity and unitarity from tree scattering, which is determined by symmetry. This is a revival of S-matrix theory, as applied to field theory, and it is a welcome breath of fresh air, after 40 years of heckling and suppression.
This post imported from StackExchange Physics at 2014-03-17 04:07 (UCT), posted by SE-user Ron Maimon