I guess we are living in two different universes. Any job ad for an ML role or ML adjacent role says Ph.d required or Ph.d preferable. Maybe it is also a matter of location. I am in Germany.
For a plain SWE role a Ph.d might be a disadvantage here too, but for anything ML related it is mandatory from what I can see.
In my hiring experience as an interviewer, 90% of candidates with PhD or not will actually have mediocre grasp on ML. It is a rare happy day when I get a good candidate. We interview for months for one hire. I got to interview candidates worldwide so I've seen people from many countries.
As someone who hired for this in general we'd use PhD (or maybe a Masters degree) as a filter by HR before I even saw them.
It's true that a PhD doesn't guarantee anything though. I once interviewed a candidate with 2 PhDs who couldn't explain the difference between regression and classification (which was sort of our "ok lets calm your nerves" question).
Yeah, you don't want to be anywhere near a place claiming to hire HS graduates/4chan posters in disciplines requiring advanced knowledge for successful product development, unless, idk, they have demonstrated mathematical talent through well-established media e.g. math olympiads, thesis on some relevant discipline.
Almost all the time, they're shitty startups, where bankruptcy is a matter of time, run by overpromising-underdelivering grifter CTOs pursuing a get-rich-quick scheme using whatever is trendy right now -crypto, AI, whatever has the most density on the frontpage-.
Yeah true, I've had to work with too many fresh college grads to not relate to this. People try to take some rare case and generalize when that's really not applicable.
For a plain SWE role a Ph.d might be a disadvantage here too, but for anything ML related it is mandatory from what I can see.