I want to do a PhD in AI/ML but the money I am earning (though peanuts) is hard to leave behind. It seems tough to go back to school, spend 4-5 years on just the survival money, not travel, commute without a car, and stay in a cheap accommodation with just the minimal things. AI/ML interests me a lot (my work overlaps with ML but I do not have any publications). I have taken online courses and studied ML books. I want to hear opinions from PhDs about their journey, experience, and if at the end it was worth it? If you dropped out of a PhD program, that would really help get another perspective, and make my decision easier.
There are many decent courses in Coursera, Udacity, and Fast.ai that you can take for free and learn the equivalent of a university course. In fact, some are taught in University. The fast.ai course has a novel teaching method (examples first, then theory), is aimed at practical coding and their solutions have. Outperformed academic papers. If you want to just learn ML/AI, this is viable. Also, ML is undergoing rapid change and most advances are only in papers right now which are all free to read online.
Yes, I want to learn more about ML/AI to advance my career, and because I like it. But I don't want no-phd to be a hindrance. If going for a PhD, doing some research, and then coming back to industry has been worth it for folks here, I would happily take that route.
Why do you want to get a PhD? PhD is for a career in research and academia. for most industrial work or industrial research labs you can get by with master's degree in DL.
On an unrelated note - a PhD by definition is not about taking courses, but pursuing you own novel ideas, and thereafter publishing them with scientific rigor. Just sayin...
Identify an open problem and go deep. credibility is peer acceptance. One should be able to do all this without PhD. The lure of your existing salary has made you a cog in the wheel.
I got my PhD in education, did quantitative modeling and econometrics for policy evaluation. Used R and SQL quite a bit. Switched to devops consultancy specializing in AWS because the work is more interesting and applied, plus the pay and quality of life are better. I enjoyed getting my PhD for the most part. I didn't go the faculty route, which sucks generally anyways, and the resaerch jobs were just meh. That said there's no doubt you could teach yourself to perform at PhD level by reading top papers and the work they cite, if you have the drive and motivation. Think of some projects to extend it and try to do them. As a grad student you'll be doing that 98% of the time anyway. Also if you have to pay for your PhD you shouldn't do it. Many places will pay you to get a PhD with a stipend, mine was 35k a year, included conference money, and tuition was covered. No teaching or TA, just 20 hours of research as a grad assistant on a faculty let project.
Can you explain - "you could teach yourself to perform at PhD level". How do I do that? There's so much on the internet that I don't want to be drowning in an ocean not knowing where to start and which way to go. My company wouldn't pay for the PhD, so yes I was planning to get into it full time by myself. Although I am still weighing my options.
What is your end goal? Do you want to create new ml and ai algos / explain how they work, or do you want to do applied machine learning on projects at a job? If you are serious about creating new algos or doing basic research on existing ones then a PhD is a good fit. If it is an applied route you're after, then I think you could teach yourself or get a master's. Give me a sense of what your interests and and I can give you an idea of what I would do to learn it.
There's no point doing a PhD and ending up again in the industry as a software engineer. I find it boring to be a SWE, but instead of doing a PhD and then again being a SWE, I could do something radically different in the same time -- say go to med school (4 years, 3-4 for most residencies).
I came to this country for PhD after getting my bachelors, had a very mean professor as my adviser who gave me so much stress and yells at all his students until they cry(including all males!) that I almost had mental issue. I remember many nights crying myself at dorm, not knowing where life is leading me to. Luckily my spouse also came to the country for his PhD in a better school with a tough but more reasonable advisor, so I got married, transferred to his school, and got my masters there then start to make money in tech. That being said, I was in EE instead of CS, and my advisor was famous for being mean in the whole university, I was also young and came from a different country, not fully understanding if I want PhD or just a school that waives my tuition. Yours could be different and easier, but in general do not expect to be treated as respectful as in your job now. Money wise I do not see a difference between phds and non phds, but Iām just regular software engineer not ml eng
Thanks. I didn't think about this aspect, definitely helpful. Do you suggest doing a thorough research about the professor from his previous students before enrolling?
I am a PhD so let me set the record straight. First almost nobody finishes in 4Y - only if they have an MS or no MS is required. Second, it's only worth going if you can go to a top 20 school. Third, academia is a 20:1 pyramid game with a dice roll at each step mixed in. If you are lucky like that chick in Deadpool 2 (Domino) then go for it, otherwise run away. I made it all the way to professor at a top 20 school but the pay was horrible. There are a lot of really bad PhD jobs out there for starry eyed workaholics who believe in winning the lottery. Most of the factors of success are outside of your control. I had all that it takes (my parent taught at a top 10 school) except timing (graduated in a bad year) and luck (my field of CS was not hiring in the year I graduated) Take classes (Coursera) or go part time to classes (non degree). AI is the most hype driven field and it's at the 25-year peak in the hype cycle. "Do you know that in 6.034 (intro to AI at MIT) they use the same final exam every year? They just CHANGE THE ANSWERS!!" is a joke from the 1990 peak of the AI hype cycle ...
What do you do at Lyft? Do you think youād have gotten the same job without a PhD too?
I see several people in really good positions directly after PhD - Andrej Karpathy, etc. I was thinking PhD would make me a valuable resource for every company which does AI research.
A PhD can open some doors to opportunities you wouldn't have otherwise. If you have concerns regarding lifestyle it's probably the wrong choice for you.
Itās a young manās game. If you are already into your working journey and plan to stay in industry it may not be worth the time and effort. You can look into enrolling part-time though, if itās for love of learning or for the degree. A lot of companies pay for it.
Part time PhD? Is that a thing?
Yes.