While there is a need for strongly typed languages, that doesn't imply that all languages should be strongly typed. More to the point, however, Scala appears to be staticly typed (I'm believing documentation here, I've no experience). Many problems are addressed only with difficulty via a staticly typed language.
Compatible with Java. OK. So is Jython, so is JRuby. Object-functional? Not quite sure what you mean, but I would guess that so are Jython and JRuby. Also Groovy.
This isn't really a response to the article, but rather to your comment. Unless you are in love with the Scala syntax, you don't seem to justify your point. Even Clojure would meet all the benefits that you list. (As well as several other languages.)
Personally, I dislike intensely Java's 16-bit char system. I much prefer either utf-8 or utf-32. Perfferable either chosen as needed. Alternatively the Python3 opaque string type with conversions to the desired representation also has its benefits. (My real preference is uft-8, but then most of what I work with is ASCII, and I only need occasional double or triple byte characters. But for that to work the language MUST have appropriate library support. As Python, Vala, D, etc. have. Ruby has it via an add-in gem. Java doesn't seem to really have it, and as a result neither do any of the languages that are symbiotes. C and C++ are, admittedly, as bad as Java. You need a large and clumsy external library. Racket Scheme has this aspect handled well, but there are other reasons that it's less than desireable.)
So. Which languages will you need in 10 years? It's one that isn't popular yet. Vala is a possibility. So is D. And prehaps there will be applications for which Swift is desireable. I'm really dubious about Java. C will probably still be necessary, but I'm not sure about C++. Some successor of the current Scheme versions would be desireable, but it MUST implement IPC much better than any current Scheme does. Some dataflow language would be highly desireable, but I don't know of any decent conderes. (The one's I'm aware of are too specialized...though one of them could grow out of that.)
The language really needed hasn't yet been written. It will be designed to be easy to write multi-process programs in. And it will be easy for processes to submit messages to each other's read queues. Erlang is almost right, but it concentrates too much on immutability, which works quite well for a certain subset of problems, and is terrible for many others. The reall concept needed is isolated mutability, where mutability is all "thread confined" (except that I mean process confined). I don't think that it should be possible to pass pointers between processes, but perhaps it could be done if the pointer only pointed to totally immutable data and it's recursive equivalents.
As I said, this language doesn't seem to exist yet, but various languages have implemented pieces of it, so I don't see any intrinsic difficulty in creating the language.