In a few areas that's true, but there are wide swathes of computer science that change more slowly, and not only in the Java enterprise space. Machine learning changes year to year, but undergrad-level foundational material from 2000 serves you reasonably well today; you still need to know what a loss function is, what cross-validation is, etc. Architectures change, but you should still know what registers are, what an instruction pipeline is, all the basic material, enhanced with some advanced stuff. As far as PLs, a grounding in all the major programming paradigms---functional, object-oriented, procedural, etc.---is still useful to have. And so on.
The big-data space is increasingly important within CS, and I don't see a big move away from academic content there. Whether you learn it in an actual academic program or self-teach, you'll need a solid mathematical background.
If you're interested in how little programming languages change over 40 years check out The Next 700 Programming Languages[1], from 1966. Sure, it's easy to point out advances like ATP or modern concurrent, generational GC, but the underlying ideas are relatively constant.
The big-data space is increasingly important within CS, and I don't see a big move away from academic content there. Whether you learn it in an actual academic program or self-teach, you'll need a solid mathematical background.