I have applied to start a PhD this year and I'm 50.

It doesn't actually matter what you do your PhD unless you plan on staying in academia. Very little of it is going to be valuable to employers, but very few employers are going to care because the criteria for getting a job in industry in computing are (a) have a pulse (b) demonstrated ability to program a computer.

> Is it more advisable in long run to stay and get your PhD or just leave and join some ML role in industry that takes Masters guys and get some real world experience on how to use ML to generate business value ?

Maybe do both? Can you find an industrially-focussed research project?