It's very much like the dotcom bubble.

It's going to burst for sure, but people will be using ML/AI anyway.

If that's your concern, then you can safely do ML/AI.

But, don't hold your breath on getting a presrigious role after PhD.

Because, there are very few real AI companies. And their hiring is skewed to Stanford, MIT, UCLA-B, Oxbridge, UToronto, ETH, etc. And getting into these schools in AI was always competitive, but now it is crazy-like because all the prep school kids are eyeing this since basically high-school or even before.

So, there is someone like me, who didn't know about modern AI until last year of college in a different major, then learning on my own and getting research jobs in small time companies with shit pay, and then there are people with tiger parents who are white collar or even academics who helps their children get into a really prestigious school, and there they do AI projects with top professors, who write them recommendations, and who also do summer internships in DeepMind, and then they use that to get a job in a proper ML company. At this point they have 2-4 publications in top tier conference. Then they work in BigTech AI lab for 3-5 years and get at least 4/5 more papers (no upper limit). And these are the people who are going to Stanford PhDs. Not people like me. And then these PhDs will be Research Engineers and such in DeepMind, OpenAI, etc.

So, before deciding to do a PhD from non elite institutions, think hard.

Because of AI hype, the situation is very bad for the rest of us. Because all of the prep school types in STEM want to make it in AI.

I'll add that I'm at out of work Software Engineer with a BS in Comp Sci and I've been researching to see if a Masters in ML/AI is worth it.

I haven't seen anything indicating it's essential, and I still see that with a MS in AI/ML you're still most likely going to be doing Software Dev... I'm sure it's different for a PhD, but as the other commenter said, it's going to take a lot longer.

Let’s just say if you wanted to have hit the LLM bubble today , you would have had to have gone into the unpopular ML field of NLP, while watching all your vision peers getting all the highest paid jobs.

So the next hot thing is likely an ML field that is not the current hottest.

How do you personally define “good idea”?

If this is about your financial outcome, make sure to factor in the opportunity cost of a PhD. It will require 5-7yrs where you will make very little money.