General answer - as hard as becoming a specialist in any other activity.
A more narrow response - depending on the purpose. If you want to understand fully the theory, then at least 5 years will have to absorb the books and then articles, and even ponder all of this. If the target - application programming, then everything is easier. Enough to know one of the most convenient languages (C++/python etc.) and to master any library (OpenCV, Tensorflow, etc.). With this knowledge it is already possible to find a job and continue to develop the theme in the process.
otha_Koepp answered on June 14th 19 at 19:19
Start doing some personal project. Surely you have a hobby outside of work. Think about where you can apply machine learning. And then study and search for ways to solve this problem.