This 120%. Your day-to-day, your mental health, and your research success are largely a function of your lab's culture, and it's the PI who sets that culture. And the unfortunate reality is that PIs really don't get any formal training on culture-setting and people management, so there are some really successful folks out there who suck at running their groups.

The way you tilt the odds in your favor for getting a good PI-student match is to be embedded in the field. If you're an undergrad, getting involved in research is a great low-risk way to do this. People in research have reputations in their field, and people in the know will direct promising students away from labs with bad cultures.

This is my advise to people who ask me for advise when applying to PhD / graduate school for research. The advisor is probably the only other main factor besides finances and any other personal constraints. If you have a great advisor all around (not just the best researcher in their field, as in they also know how to mentor, teacher, network, and help you reach your goals), they will know how to navigate your different research interests and goals of your PhD and get you to where you want to go.

I think the entire grad school "industry" really needs a "Me Too" moment. Abuse and exploitation are the norm it feels like :(

From my experience

1. Don’t opt for a phd just because it sounds cool (remember the first tweet shared in that article).

2. Academia is a pretty hard space if you can’t get your own funding or if you don’t like hustling for it. Getting tenure track jobs without having funding attached is impossible.

3. Very few jobs in industry need a PhD. Most are ok with a masters. Those that require a PhD can be selective based on where you graduate from and what your publication record is.

4. Know your PI well and reach out to his/her previous students who have already graduated. Some may reply - especially those with a bad experience.

5. Back to point #1, don’t take up the PhD just because you get the assistance and stipend. Make sure you really like your subject - it’s the only thing you’re doing for the next 4-12 years depending on your discipline while subsisting on poverty level stipends

> Very few jobs in industry need a PhD. Most are ok with a masters.

This varies dramatically by field.

What industry has a heavy concentration of PhD except academia?

Not the GP and I agree with your broader point that a master's is sufficient approximately 99% of the time, but a few places come to mind: SpaceX, NASA, CERN (is that technically non-academic? I don't truly know), probably certain departments of US Dept. of Energy, things around nuclear power.

They probably don't have high concentrations absolutely (e.g. they're not going to be 90%+ PhDs) but it's probably still several orders of magnitude above your average industry role.

IC design for behemoths like Intel/AMD/Nvidia. AFAIK none of those is gonna let you anywhere near their multibillion-dollar design without being at the top of the field or having lots of experience (>5yrs) already

Is 5 years really the bar for "lots of experience?" I feel like this is a byproduct of current (last 10 years or so) frontend/bootcamp dev mentality that you're a senior engineer after a 6-week course and 2-3 years of experience at a consulting body shop.

Even if you get a PhD your working career is going to be around 35 years. Add 5-10 if you're done after college, and even more if you're not doing college.

I think we need to stop pretending people are senior when they're 10-15% of the way through their career.

Robotics is one such field. I work with tons of PhDs, and self driving car companies hire them like crazy. In certain organizations, you need a PhD to advance as an Applied Scientist. A masters won't cut it.

But that’s my entire point. Taking your example, the bulk of robotics companies aren’t all applied scientists (for that matter in spacex, nasa, etc.). You will have a very small cohort that does the r&d and the rest of the folks executing on that.

The r&d space is really interesting but very selective for obvious reasons. Just a PhD won’t get you in. You will need significant other contributions - eg the right publications in high impact journals, possibly internship with the companies themselves, the right advisor/PI that collaborates with the company, etc.

The team I'm on is half PhDs. We aren't doing fundamental research, we are building a real product. We prefer to hire PhDs as applied scientists (which we distinguish from research scientist) because they are better able to apply and build new-ish ideas from academia. But most of what we do is not novel in the publishable sense. The PhDs I work with are certainly all smart, but aren't the top echelon you are describing.

Any job that is highly research oriented is going to appreciate a Ph.D.

Whether it becomes a practical requirement, or formal requirement, would depend. If it's a high paid position, in a small research group, of an established field, then Ph.D.s are likely to be very important.

If its a position in a large research group in a new and fast changing field, then its likely to be more open to anyone with unusual or interesting promise, but still highly value Ph.D's.

As for fields: research in engineering, finance, economics, device physics, chemistry, biology, pharmaceuticals, etc. Anywhere the research involves highly technical knowledge, and a pre-existing track record of original thinking, personal initiative, that have produced objective results, are very important.

"Research" is a thought and project management meta/leadership skill, on top of being highly educated and talented in an area. Ph.D.'s provide the extra time and social context to fill out those skills and demonstrate a track record.

That's true in biology, chemistry, and physical sciences, but less true in computer science.

In any of those fields, and computers science, there is a spectrum.

Deep research areas in computer science would be foundations of databases, cryptography, formal verification, statistical heuristics, core algorithms behind deep learning, ...

Since so many areas in computer science are relatively new, there is generally more flexibility in who might be considered great at research. The field includes many unusually creative individual practitioners who have produced great work without a Ph.D.

But Ph.D.s are still very much in demand and common for research positions in these highly technical, high value areas.

Yes - this is unfortunately everything. There are a number of categories to optimize (research area, lab size, funding, university, other lab members, etc) but the biggest factor is your relationship with your PI. I cannot imagine working as a PhD student in a lab where I don't deeply respect and get along with the PI. I was very lucky but selected for this when choosing PhD labs. A surefire way to have a terrible few years is to ignore others advice about PIs to avoid.

I chose mine because he was a reasonably guy to work with and talk to. What I did not know at the time was that he was very junior in the department and needed to please other professors. So even though I liked working with him, I was quite helpless at certain difficult parts of the process.

Far too few incoming students value this, even if explicitly told. They will go to the lab of the most famous professor at the highest rank university, even if that PI is not a great mentor. Not all PIs are intentionally aiming to make their student's lives hell, but many inadvertently do by providing insufficient mentorship, which results in students suffering huge amounts of anguish because they are struggling to excel and know it.

If a PI isn't meeting with each PhD student for at least 30-60 minutes per week, especially in the first few years, they are probably not doing a great job.

I always encourage students to email the current and former students of the professors they want to work with once admitted to see if they can have a video chat to discuss what it is like working in the lab. Even then, it can be hard to get frank feedback.