I'm about to graduate from a (non-US) master with a strong emphasis on theoretical stuffs (automata, logic, type theory and so on) and a pinch of ml. The topics were fascinating for me until recently when I had to make a decision on a tcs phd offer and realized that it potentially causes a career suicide given that
1. I will probably never get a Turing or become Gödel and, somehow consequently, have no motivation to transform my interests in researches to interests in building a career as a researcher.
2. I got zero experience in the industry and will remain more or less so until finishing this phd in my 30, which potentially confines me within the highly competitive academia. I was stupid enough to believe that the only hiring bar for FAANG and alike is the elementary icpc stuff and whoever played with serious algorithms and a handful of languages and tools can pass it.
3. People are practically re-defining cs := machine learning and cs_why_doing_it := cs, which used to make me cringe. Now I start to agree on them since I find the latter one more like stuffed with random adult toys crafted by theories with few intrinsic values in it apart from letting academics pleasing each other, instead of a tiny brick on the holy road towards the discovery of logic and nature like I used to religiously believe.
But still,
1. It's a phd and phd is a thing.
2. I don't hate the topic and there are still reasonable amount of connections with the industry (focused however still only on the most formal aspects).
3. It's corona time and I could end up with nothing if not accepting this offer on time.
4. I could always quit, though not sure if it's a good idea.
I'll wholeheartedly appreciate it if someone could forgive my ignorance and save me from a tangible mental breakdown with any advice.
All things equal, it's generally good to give yourself options. If you've never experienced working in industry, what if you got a job in industry for a few years (ideally at a few different organisations, since there's a lot of variability), to see how you found it? Getting a broader range of experience may help give clarity.
If you're concerned about being left without something to do in the short term, what's the harm in starting something but keeping actively looking for other better opportunities in a fraction of your time? If you find a better deal you can drop what you're doing and switch. I used to work for a small business that took on a few applied math PhD interns, one of the interns found the commercial side of things much more engaging and dropped their PhD to switch to work full time. Conversely there were other employees who switched from full-time work to part time so they could pursue masters or PhDs.
Partway through my PhD I realized I had no interest in being a researcher. The reason was that after seeing what life as a tenure-track professor and/or researcher (e.g. Postdoc) was really like, I had no interest in that life for myself.
Forget research interests, being the next Godel, and all that. Ask yourself, is life as a researcher or professor what you want - because that is what you will have to deal with no matter what your research interests are or if you are working at an R1 university or community college.
Also, life is long. Don't worry about making the perfect career decision. Just make the best decision you can.
The world is actually built by nameless people like us. The legends are simply symbols representing the work of people before them, not goals for ourselves.
Your work has value. Don't ever let anyone (read: employers and yourself) belittle them. Employers devalue your work so they can pay you less. You devalue your work because you compare it with the accumulated work of thousands of people across generations, which is downright silly, don't you think?
As far as actual research topics, there's a long standing need to integrate symbolic logic into deep learning models. No one knows yet how to do this. If you have an interest in the intersection of tcs and ml, I'd encourage you to work on that problem. The next few decades will be known in human history as the time when AI emerges. You decide whether you want to watch it happen or help make it happen.
I've got a PhD and I don't agree with this at all. People get PhDs for all sorts of reasons, including not knowing what else to do in life. I know a lot of PhDs and I'd say only a few had "passion" for what they were doing as grad students. And for some of those passionate people that passion was driven out of them by graduation time.