Very interesting proposal on selecting Programming language for Tensorflow by Google team. Swift or Julia for TensorFlow
One of their project goals is partially achieved by Julia now,
Debuggability and Introspection: An important part of R&D is figuring out what is happening in a model, diagnosing failures, and figuring out what to change. A system that diagnoses more errors earlier (e.g. shape mismatches at compile time) is more usable than one that only catches them at run time.