Should You Go to Grad School for Artificial Intelligence?

is it worth it to go to graduate school for artificial intelligence find out in this video but first if you're new to the channel I am dr. Phil Taber I have a PhD in condensed matter physics that I got in 2012 and probably went to work for Intel Corporation as a back-end triage process engineer I left in 2015 to pursue my own ventures and have been doing freelancing consulting artificial intelligence ever since quick note if you want to get a jump start on learning how to read deep learning papers and turn them into code you can check out my new course deep Q learning from paper to code on sale at you to me right now for $9.99 but back to the topic of the video so when you're thinking about going to graduate school you really want to ask yourself a few different questions what is it that I hope to get out of going to graduate school what is it that I'm going to be giving up by going and what is the probability of success in achieving my goals so if you are looking to become a researcher in the field or a faculty at a university then you pretty much certainly have to go do a graduate degree in artificial intelligence computer science some related field now if your goal is to get a job a well-paying job then it's a little bit more murky so there are many companies that will hire you without a degree at all and many for which a portfolio plus an undergraduate degree would be sufficient heck in some cases you don't even necessarily need to know an undergraduate degree you just need a strong portfolio I recently saw an interview with a guy I forget his name Emil something who landed a job as a Google no less I believe as a self-taught developer he went on a mission to find himself for a few years after graduating high school and then decided to become a deep learning engineer and did quite well for himself on the self-study track so you may not even necessarily need an undergraduate degree if all you want is a good job at a great company so ok let's assume that you want to push the envelope and really push the limits of understanding of artificial intelligence or you want to be a faculty on a university and so you pretty much have to go to graduate school so what is it that you can expect when you get there well now full disclosure as I said my PhD is in physics however in conversations with other student and looking over other programs in various fields the experience is more or less Universal the only thing different is what you're studying so there are a number of common factors to which I can speak on and in particular I'm thinking of of course the coursework you have to deal with but this is just the easiest part I'm making a mistake the coursework is difficult and you're gonna be tested to the limits of your understanding and that you know the homeworks will be difficult the exams will be difficult the whole thing will be hard but that is in fact the easiest part typically following the coursework you will have some set of written qualifying examinations now the written qualifying exams you don't get any notes usually with software engineering or computer science it may be a little bit different where you have you know take home tests or something like that or it's an in-class exam with a computer and you you know type out solutions to problems you know through code I'm not entirely certain but either way you're going to be timed you're gonna have a set of problems you probably haven't seen before and you must pass it in order to become a what is called a PhD candidate so if you pass these exams you have completed generally the first step in becoming a PhD candidate and there is in fact an entirely separate step to completing the process of PhD candidacy so the second step is in many cases proposing a research topic and my graduate program it was simply presenting a sort of defense of someone else's paper wasn't really necessarily related to what you want to do although it was generally tangential but we had to go and prove that we could dissect a paper understand the arguments presented find the weaknesses in those arguments where the author's may have left some holes in their argument that perhaps the reviewers didn't catch you had to demonstrate a capability to think critically think scientifically and to generate new ideas in research so it's kind of a mini defense but on someone else's research which makes it even more difficult once you pass the course work written qualifying exams and your oral defense of your PhD candidacy you become what is called a PhD candidate now at that point you are probably done with your coursework maybe you have one or two left to take more advanced courses that you weren't able to stuff in your first couple years but at that point you are full-time research now this may sound awesome but this is actually the most challenging so what can you expect while doing full-time research as a graduate student so everybody's experience is gonna vary and the probably biggest factor in your experience as a graduate student is going to be the professor that runs your lab now when I was at West Virginia University I worked under Sergey who was a hardcore Russian brilliant scientist one a man I look up to even to this day as a you know incredibly competent intelligent human being who taught me an enormous amount really one of my first mentors and he was really great to work for it on what's kind of funny is that other students had warned me away from it because he had fired another student a couple years prior to me getting there now he had fired the guy because and I saw this with my own eyes he went to work for another faculty and every time I would go down there to use their equipment the dude was farting around on the internet looking stuff looking a cute cat videos not doing anything related to physics generally not a very ambitious or driven student and so the guy got you know fired from the lab because my professor was a new faculty at the time which is a big influence in my experience as a graduate student so as a new professor they are had to hit the ground running they are expected to produce something relatively quickly the mantra publish or perish is very much a real thing and it is the reality for these faculty you have the publish to not only get funding but also to meet the requirements of your department you have a department head who says you got to publish X number of papers per year and they enforce that pretty strictly and as a new professor you can be fired not at any time but you will have a review coming up and if you didn't meet the requirements of what they set forth you're out of a job and you got to go find one somewhere else and keep in mind there are only a limited number of these positions there's if you want to be a faculty you're gonna be competing against 100 to 300 of the people for one position you know there may be 10 20 positions open there may thousands of people applying for them all of them applying for the same positions so you're gonna be in fierce competition and even when you get the job it is far from certain that you're going to keep it so in my case being the only graduate student or at least the first graduate student he hired some others as we progressed along but being the first graduate student in there meant I got to work very very closely with my professor who was a total workhorse the man worked 14-hour days I'm not kidding or exaggerating he would leave it at 2:00 a.m. and be back at 8 a.m. to teach classes the next morning pretty much seven days a week both teaching classes on the weekend but he was still working you know 12 14 hour days and so I got to work very very closely with him and really learn the ropes of scientific research by watching a professional do it which you may not have the experience now my wife whom I met in graduate school worked for another faculty who has already established had a lab with many students and as a had a totally different style of managing things now I worked very closely with my professor and anytime we ran into an issue we would problem-solve it together he would help guide me for the most part some stuff I'd to solve on my own but there was certainly support if I needed it in my wife's lab it was totally different the students were completely and totally on their own and this is actually the norm if you were working for a well-established professor they are they're not lazy it's not like they don't work it's simply that they have higher level tests so kind of like a CEO their job is to bring money into the department they had to bring in millions of dollars to the department because the university system is horribly bloated it's a huge bubble but that's the situation we're dealing with and the consequence is faculty are pretty much not obsessed but have to work tirelessly to get funding to secure funding to give talks to network to make new connections to find other people to do collaborations with and so they have a full-time job basically selling their the research of their lab just like a CEO asset sell their product so the consequence of that is that you the student are gonna be left on your own and in fact you're gonna probably be relying on other students or maybe postdocs for support and help but remember they're in the same boat they are trying to get their PhD or the postdocs or even worse off because they have the mentor the students but they're also trying to become a professor as well so there is a an enormous constraint on intellectual capital in these well established labs where students are kind of siloed on their own individual projects you may not even know enough to help the person next to you you know about their projects because all the projects are heavily specialized that's what being a PhD means is getting you know really really deep into one thing deeper than anybody else in the planet becoming an expert in and presenting final research but that just means that nobody really help on the detailed nuances of your problem even if they want to and they may not always want to so there are situations in which students may actively try to sabotage you and I dealt with that as a graduate student myself so in my fourth year I started one of the benefits of my lab was I got to kind of do different projects I was I was given a lot of freedom to do kind of what I wanted long as it was you know something that we could publish and you may not have that experience but that was mine but I worked with this other tyrannical professor that my wife worked for and we worked in something called molecular beam epitaxy it's a way of growing thin films it relies on ultra-high vacuum so atmospheric pressure is like 700 and the pressures we worked at were 10 to the minus 12 so or 10 to the minus 10 sorry so 12 orders of magnitude difference they're ultra ultra high vacuum ultra pristine ultra clean environments and their giant chambers you know several feet in diameter four or five feet tall with really expensive vacuum equipment hooked up to them so the particular type of vacuum equipment is called a cryo pump it goes down to 4 Kelvin which is you know really really cold and basically what that does is freezes out all the contaminants in the chamber so anything that bounces up there because you're the molecular flow regime get stuck on the carbon and basically removed from in the chamber and so there is no contribution to the pressure now what this means is is that if that pump heats back up all that stuff gets you know belched back into the chamber and you end up with a contaminated chamber that has to be opened up and then shut down and then pumped out with a manual pump and then baked you have to bake it with a heater you have to coat it in a big blanket and then you know like it's cold and warm it up over the course of a week it takes a week to go from atmosphere down to 10 to minus 10 and so what I found in my last year my fourth year at WVU was that I would come in always on a Saturday and my cryo pump would be off this happened several times only on a Friday night none of the other chambers had the same issue even though they were all in the same circuit it wasn't a power surge it was almost certainly human intervention another student I believe who was jealous of my relatively modest success was sabotaging my efforts at research but I didn't let that bother me you know I just you know took it as a sign of whatever I got a week to do to read to learn to do other stuff it wasn't like every week it happened it was periodic is two or three times and as soon as I left my professor got a job at another university I followed him there as soon as I left the problem stopped it never happened again so it was certainly a human intervening to trying to hobble my research and you may find this in graduate school yourself unless you have a personal vendetta though it's unlikely but you know the more competitive environments that may be a common practice with students actively trying to sabotage each other because remember you're all in competition you're all gonna be going for postdocs to are only going for a limited number of positions it is a very very cutthroat and if you do get a faculty position you're competing with these people for you know finite funding dollars so another thing to consider is that they may not be able to help you they may not want to help you in fact they may want to actively hinder you now other things that aren't talked about about graduate school are the you know kind of the emotional rollercoaster side of it it is an intensely rewarding experience to see a novel result for the first time to see something that no one else has seen particularly in physics when you're getting kind of deep insights into nature how nature works it was quite an exhilarating experience to see it and it would made you want to work more made you want to work harder to see no to probe the boundaries to see how far you could take the effect you know to see all the different things you could learn but that came at a significant cost many many long nights many many nights spent alone doing research so isolation is relatively common you know students are siloed faculty are siloed to a lesser extent but a little bit so isolation is one common emotion I didn't experience it too much but there were many nights in which you know it wasn't the norm but there were many nights in particular on isolated events where you know people are out having fun so those students weren't as as dedicated to research as I was so they were out partying but I couldn't go because I was running an experiment in the lab and it required you know it couldn't be automated I had to be there observing in real time to see what the sample was doing and so this may be the case with you as well and in fact I heard from many individuals to varying degrees some worse than others I knew one student who whose project didn't work and that was quite common and in that lab if it didn't work he were screwed you know they weren't the professor wasn't going to shift you over to another project because the grant funding was for that project in particular so he wasn't kind of uh he didn't loosen that restriction and allow students to kind of help each other and work on different projects so that way he could maximize a total number of papers he didn't care he was already well established didn't really matter to him it was on the student to get it figured out didn't matter if if the physics behind it was wrong and it would never work even in principle that didn't matter only matter there was a grant and it had to be done and so the student was left holding the bag at the end of the day and the consequence was a dude ate himself to medicate you I'm sorry he didn't eat himself he wasn't cannibalizing myself he ate too much to to medicate the issue away so he tried to drown his sorrow in food and you know we know then result of that you know and it's not a you know not something you really want to experience of course not something you have to do other students you know had more positive reactions to it they hit the gym they just worked harder or they found other ways of solving the problem but just know that is an emotional rollercoaster where you have intense highs of seeing new things or if you're an intensely curious person if you like research you're gonna get that dopamine hit from you know seeing something new but the flipside is you're gonna have to sacrifice your social life probably you're going to sacrifice human connections and it could at times be lonely and isolating and in some extreme cases if you allow it to go that - you know that far if you don't take ownership over your situation and your emotions you could slip into depression and anxiety but that's only if you choose to you know I believe you have the choice whether or not to do that you could you know take charge of the situation and improve your life in other ways so once you have finished your research your next order business is to defend your PhD now what this entails is you must have already written at least one paper typically you know two three four five I have five on the first author and I think two or three of them two of them I believe and then co-author on three really two papers first author is really what you're looking at minimum some students do better than that others only do one and just get out out of a you know out of pity that we want to get rid of you whatever you're clogging up our pipeline you're clogging up our numbers some students get out you know on on pity basis but don't count on that and you really don't want to be in that boat because that takes seven years it only took me five I was on the lower end of the distribution most people was five to seven I was ambitious and kind of worked hard from the very beginning so I got out in five years luckily but once you have published the papers you have to write up your dissertation which will be kind of a recap of your papers as well as the background in the field and you have to choose your PhD committee usually which is the people who will judge you and determine whether or not you get your degree choose wisely choose very wisely and you know it's kind of interesting I picked people that I thought would be not soft on me but would ask the right intelligent questions and kind of challenge me but not try to take me down you don't want someone to try to take you down in fact one of the faculty did do that it was quite strange my my advisor wasn't there because we had switched universities and he was still I think we had to have someone else chair my my committee but of course my advisor was still you know there via Skype helping me along so I was kind of alone with the wolves but he was there listening and offering in you know what help he could because he had a vested interest in getting me out and graduating his student of course but it's a very intense process very intellectually intellectually rigorous they have read your thesis they will read your thesis they will find the mistakes in your thesis they will make you fix them and they won't let you finish until it is at a state in which they are happy with it and that you know how long it takes is up to you and how well you write so it could be good it could be bad but you choose your committee wisely once you pass that you are done you are finally done and hopefully you spent the last six months looking for a job because that process takes time to all processes take time but at that point you are free to go free to proclaim yourself a PhD which is you know a pretty good feeling to be amongst a handful of individuals that have the intellect the drive and the opportunity to do so so you know if you do it if you've done it now that's quite an accomplishment I salute you for that and just keep in mind that don't you know don't be you know don't be too arrogant about it be humble learn from others around you particularly those with real-world experience and do your best to work hard at your new job at least as hard as you did on your PhD so with all that said would I go back and redo my PhD experience now I wouldn't go now at almost 40 there is no way you could ever convince me to go back to school for anything for any reason I'll take a few online courses if that but there is no no universe in which I'm going back to school I'm glad I went and you know at the time I did and I certainly wouldn't change fad it was an intensely rewarding experience he was challenging certainly I'd push the limits of my understanding push the limits of you know what you can tolerate mentally and emotionally but at the end of the day it was a positive experience I came out of it a better person better researcher I came out of it slightly smarter I hope so for me it was incredibly worth it though some others may have had a different experience and I wonder to what extent that depends on what they put into it I'm a firm believer you get out of life what you put into it and some people just didn't put very much in so I don't expect them to get very much out so I hope that was helpful let's recap a little bit why should you go to graduate school you should go if you want to become a world leading researcher if you want to become a faculty if you want to lead a research lab then you certainly have to get a PhD there's no two ways around that but just know that you're in for a rather long five to seven year experience where you're gonna have to take and pass classes which is the easy part you're gonna have to pass a written exam you're going to pass some sort of oral defense of either your topic or some other topic related to your field and then you've got to publish papers you have to do the research and oh and one of the thing I didn't mention is you're gonna have to teach other students you're gonna have to be a TA which means teaching labs usually and grading papers that's a couple years of work that's a nuisance but it's not a big deal we all did it it's you know it's okay not not a huge issue but it's something you got to do is you know part of the trial and then you will finally get out and you'll find a job hopefully waiting for you if you're if you did your part published well made connections applied to lots of jobs before you got out so I hope that was helpful if you found it so please make sure to hit that like button leave a comment down below with your experience in graduate school if you've been I would definitely appreciate that I remember my new course DBQ learning from paper to code where I teach you how to read deep reinforcement learning papers and turn them into pythonic pi torch code is on sale I bought until the end of the month at $9.99 I look forward to seeing you in the next video

Channel: Machine Learning with Phil

Views: 23,087

Rating: 4.8702064 out of 5

Keywords: should i go to grad school, should you go to grad school, is grad school worth it, is grad school hard, should i get a phd, should you get a phd, is a phd worth it, is a phd hard, how to deal with loneliness as a phd, phd defense, how to choose a phd committe

Id: qiH6pJkspMs

Length: 20min 25sec (1225 seconds)

Published: Wed Jan 22 2020