If somebody wants to learn now, its good to go after the future market which would be dominated by mobile, apps on cloud.The (now) largest mobile platform uses Java as its main language for applications. Lots of cloud services, including much of e.g. Google's, are written in Java.
(Scala, Clojure seems to be picking up)
We live in a bubble:
http://www.google.com/trends/explore?q=java%2C+scala%2C+cloj...
The average programmer will never write Clojure. Maybe Scala, but only as a Java with less boilerplate.
Golang/Dart will be a deadly combination as well
This almost seems an enumeration of today's hypes ;). You know that Go (the language) is practically a Java 1.0? What is so deadly about a Java 1.0-like language with green threads (which have been around a long time) and CSP-like channels (for which Java implementations were available). What's 'deadly' about it?
I like cutting-edge languages such as Haskell. However, historically, it's mostly boring languages with industry support that win. C was boring and it had UNIX, C++ started out as C with objects and had Microsoft and many others, Java was boring (a watered-down C++) and had Sun, IBM, Oracle, etc., and C# was boring (a Java clone) and had Microsoft (arguably, it did adopt more outside influences than Java ever did). Go will probably be popular as well, it's an incredibly boring language (no one could claim it to be revolutionary after Java or Erlang) and it has Google.
Note that I don't find the predicate 'boring' offensive. These languages are well-understood (except maybe C++), are standardized, have large ecosystems, have great tooling. The result is that they allow you to focus more on the actual problem than reimplementing basic libraries or fighting tooling.