Languages look similar, but Python value resides in the vast amount of libraries available to do data processing and ML. Most of them backed by C implementations under the hood.
I don't do any ML and don't know much about it, but the Elixir / Phoenix people around José Valim now build ML software (nx, livebook) and I'm excited for this.
So whatever headstart Python might have in machine learning, it is only a matter of time until all major languages support the same C, C++ and Fortran libraries.
The only things holding it, are having anyone that cares enough to write those bindings, and the culture of which languages are accepted when submiting papers.