What sort of opportunity it is depends on how much you want to focus on it.
If you want to be a regular programmer, you might take the time to learn a high level interface for some machine learning tools, such as Tensorflow or Keras. There will be plenty of things you don't know how to do (even within those tools), but you may be able to apply predesigned model architectures to problems. The models won't be as good as one designed specifically for the problem, but it's one more tool in your toolbox, and it's possible you'd be able to get some useful results occasionally without devoting a huge amount of time to mastering the techniques.
But if you want to really focus on machine learning, at the research level, you can potentially tackle problems that existing techniques haven't been able to solve. This is where most of the big projects that you've probably heard of will be happening: self-driving cars, AlphaGo, etc. What you can expect here is a lot of hard work. You will need to develop a fairly deep understanding of the mathematics involved so you can visualize (to some extent) what is happening in the potentially high dimensional spaces involved, identify potential failure modes, and identify models that won't fall into them. It involves a lot of trial and error, failed attempts and gradual improvement before you're able to develop a model that beats the stuff already out there.
It's very rewarding work if you enjoy it. There are well-paying positions in the field, but that's just a bonus if you already enjoy the work, and it isn't enough of a bonus if you don't. In my opinion, going into this for just the money would be a mistake. There are almost definitely other jobs that pay just as well but don't take anywhere near the time investment to become (and stay) competent at them. But if you really want to work in this field for its own sake, and also want to make sure you don't starve while you're doing it, it's absolutely worth it.
