The limits of ML (as we know it) is still not fully known, so it's worth throwing a wide range of problems at it to see what sticks. I think that's why you see talk of ML everywhere, because people are still feeling out the space, seeing how far ML can go. Eventually, the dust will settle and the next big thing will come along.

So yes, should start worrying?

A lot of the modern algorithms we write have to do with pattern matching (eg patterns in numerical data, pixels, etc). Machine learning is basically a meta technique for pattern matching. It should probably be in your toolbelt, even if it's not your main field of expertise.

ML is delivering a lot of value, but is largely being applied to high value areas since the start-up costs of adding ML to anything are pretty high compared to traditional dev work. So it's not going to replace traditional dev work any time soon, but it's​ certainly going to be important.

I don't think it is the answer to everything, but it can be very broadly applied to a whole lot of scenarios and get pretty good results for the most part. I don't think you should worry, however you should take a look at it because it is pretty cool!