博士生的磨練。關於博士生生活的討論 (The Ph.D. Grind: Candid Discussions About Ph.D. Life)

字幕列表影片播放

>>Presenter: So Philip joined Google last month and is working with Peter Norvig building
tools to improve the teaching of programming and computer science. He's received degrees
from MIT and a PhD from Stanford just recently. So I'll turn it over to Philip.
>>Philip Guo: Great, thanks Jaclyn. Everyone can hear me from the back right? There's no
clapping involved here. [Laughs]
>>Philip: Guo: Okay, anyway. I'm gonna just basically in this hour my goal is to be just
maximally useful to like everyone in this room. If it ends up with me talking for most
of the hour that's great, if you're all talking and asking questions that's also great because,
yeah, whatever I can do to be most useful to you today that's basically my goal for
the hour. So, the only notes I really have is this [inaudible] sheet and I'm just gonna
read off the Meta stuff and then I'm just gonna start going. So, basically the format
is I just have this one thing of HTML here and these are the 20, you know, lessons I
summarize in my PhD book and this is gonna be completely non linear so whatever people
are, you know, whatever the conversation diverges towards, you know, just for the sake of the
recording, the fact there, I'll just highlight that. So, this could be any one of these 2
to 20 combos, we can just talk about them. So, whatever direction stuff is going in,
again, this is very exciting cause I have no idea what's gonna happen. So, just to start
with, like I said, this is not a typical tech talk. There was a little surprise there it
was on the tech talk calendar or something cause I actually gave a tech talk here last,
uh, last year about my research and it was a very standard scheme like, "Here's my research,
here's the slides" and then I get a talk and, you know, you say that people can interrupt
with questions but people usually don't interrupt with questions and you give a talk for a half
an hour, 45 minutes and then some people ask you questions later and you respond to some
you can't answer. So I'm sure all of you, mostly intern students, have heard quite a
few people speak in this [inaudible] and it's like I'm gonna show off what I did and then
maybe ask some questions and then they're like did you consider doing this and we're
like no we didn't. [Laughter]
>>Philip Guo: So, this will not be like this. This is, I'm highly interruptible so like,
there should be mics going around or something but like, seriously, totally interrupt me
anytime. Is there an interruption right now? [Pause]
>>Philip Guo: There should be other mics everywhere else. Anyways, so, um, I'm here as a peer
and not as any kind of authority figure or anything. So I was in this audience, exactly,
last summer. I was an intern here. It was the final, almost the final year of my PhD
and I was an intern here at Mountain View and Jaclyn and others organized a bunch of
these great talks from Google leaders, right? So, there were vice presidents, research directors,
everyone talking, you know, to interns about how great Google is and about leadership and
about career building and everything and I thought it was some really interesting talks
but I always felt that I would like to hear more from people who are in my same position.
So, I, I, you know, when I first got here, one of the things I really wanted to do was,
since it was in the summer and there was a bunch of interns there and I had just been
in the audience one year earlier and I wanted to give a certain kind of talk like that.
So, um, so, this is kind of the talk that I would have maybe wanted to hear when I was
an intern here. So I'm gonna have to do a cheesy quick show of hands, this is the only
interactive thing I promise, so how many people here are full time? Oh, we've got quite a
few people full time, okay, cool. And how many are interns?
[Laughs] >>Philip Guo: Good. Okay, awesome. And out
of the interns, how many are currently undergrads or masters students, undergrads and master's
students? Very good. And how many are doing their PhD right now? Oh, wow, we've got quite
a lot. Okay, perfect. And these will get slightly more embarrassing, so who has just finished
their first year? [Laughter]
>>Philip Guo: Okay, you see where this is leading. First year of PhD, sorry, first year
of PhD finished? Okay, good, then second year? Just finished two years? Okay and then just
finished 3? Okay, faces are looking more and more [inaudible]
[Laughter] >>Philip Guo: Just finished 4? Um, just finished
5? Okay, I'm not gonna ask about the rest. [Laughter]
>>Philip Guo: Okay, cool, so I just wanted to calibrate because everyone probably gets
different stuff out of it depending on what level you're at. Oh, actually, out of the
people here who are not currently doing PhD, how many of you might consider doing one in
the future? Oh wow, that's quite a lot. Okay, great, awesome. Cool, so I'm just gonna give
a bit of background about my PhD project and we'll just take it from there. So, just a
little background on myself, I did my PhD in computer science, it took about 6 years
to finish and I did not know I was, you know, I did not have the green light to graduate
until about 3 months before I graduated. I mean, I was working all the way to the end
and once I got the green light everything was good. I just scheduled to defend and then
write dissertation and everything checked out. But up until January 2012 I did not know
that I was gonna graduate and my date was still uncertain and I'm very lucky that everything
worked out very well in the last 6 months. So, toward the end of my PhD right when I
was around hopefully getting the green light I thought that, you know, that my dissertation
at most 5 people would get because it's basically that I convince this committee and maybe one
or two other suckers who I sucker into reading it. The papers I wrote, maybe like 50 might
read those total, right, that's like a reasonable number. But I wanted to, you know, I wanted
to have something that would be lasting beyond my PhD for this particular research. So, I
thought about the fact that, I looked at, throughout my PhD I've done a lot of writings
on my website about the expense of going to grad school and I've been reading a lot about
it and one thing that did not exist was a comprehensive account of one person's PhD
expense. There have been, say like professors that write about, oh, like, there's like,
I'm sure you all know that there's a few professors that have written about PhD advice and stuff.
Like top ten tips to getting into grad school or advice for getting your dissertation etcetera
or whatever and there's like a small set of that. So on one end there were people in authority
who were writing, in fact, wistfully about their experience and like, "Oh you guys should
work hard" and do well and stuff like I did and then the other end there were the super
whiner trend, so there were the PhD comics types of like grad students who might have
dropped out and they're like, "Oh my life sucks and I'm gonna rant about it" so if you
look up like PhD life or whatever, on Google you these 2 polar options; you get people
who are offering stately advice, you know, professors, research leaders, you know, perhaps
some of the people you've been hearing talk here and other places, you know, why are PhDs
good and so forth. And then on the very other extreme you have people who are
[Pause] >>Philip Guo: Is this working? Okay, good,
was the other one going in and out? Yeah, okay, sorry about that. So on the other end
you have, maybe it's my moving with the mic is that it? Okay, I'll try to hold it super
still; actually let me just put it on the podium here. Does this work better? Yeah,
I have to like lean in, it's weird, okay, I will try not to oscillate the mic too much.
[Laughter] >>Philip Guo: So, on the other end you have
people who are, you know, who are still going through their PhD that may be having issues
and such and I felt that neither side were able to write as sort of, you know, objectively
about their experiences. Also, there's this format thing, there's never a kind of full
account of their experience. So I decided, you know, that I wanted to be the one to write
about this. So I started accumulating notes about 6 months ago, you know, about the 6
years of my PhD and how it progressed and collecting old emails and notes that I had
taken and old papers and drafts and everything and after I finished my defense on, you know,
at the end of April I sat down for 2 months and I did this writing project. It took about
one month to write and then I showed it to some people, got some iteration, got some
drafts, took about another month to revise. And I released it online on my website and
I wasn't like intending to make money off this or anything and it just like, it was
just like a thing I was wanting to do. And I released it on my website, you know, put
it all up there and I was just like planning to email some friends and just get people
I know to read it and such. But, it actually got on to certain social media sites and then
kind of kept on spreading and by about like 2 or 3 weeks time there had been like 50 thousand
or so downloaded and I was getting a stream of dozens of emails a day about it from all
sorts of people, maybe even the most unexpected people. One, I have some of the sample emails
listed. One of my friends who was in grad school, she actually accepted full time here
but she's not here yet, but she said, "Thanks for writing this because I showed it to my
Grandma and she like loved reading it because she had always been bugging me about you know,
what is doing a PhD like? And now I don't have to explain it to her cause you just did."
[Laughter] >>Philip Guo: So, I got those emails to other
emails from like professors, from current students, from perspective students and so
on. So, I felt like it kind of really struck a nerve with quite a few different types of
audiences and, so, that's why I'm here today talking to all of you. And I don't think it's
a great use of time for me to rehash the contents of my book because it took me several months
to distill down 6 years into, it's only like a hundred pages, it's very short. Like, for
people that have read it you probably read it in like 3 or 4 hours or something, it's
actually quite short. But I don't think it's very useful to distill that down to like 20
or 30 minutes. So I am just, I would just be thrilled if there were questions or there's
just things we could launch off talking about. That'd be great. So, otherwise I'm just gonna
go down the list. Um, yes, in the front.
[Inaudible]
>>Philip Guo: That's a good question. Um-- [Inaudible]
>>Philip Guo: Yes definitely, yeah, definitely. So the question was did I take on any under
graduate minions while I was doing my PhD? Is that the question? Um, so early on in my
PhD I was, you know, I was on this big group project that people who have read it read
about and there was one or two undergraduates who were working with us also but the project
was so disorganized that like, it was like the blind leading the blind, like I was a
first year student and then my advisor wanted me to lead some undergraduates to do stuff
and it was not a great experience for either party. And later on in my PhD I kind of, I
managed to do, I managed to pick projects that were very self contained so I did everything
myself. So I don't actually have the experience working with undergraduates personally. Sure.
[Inaudible]
>>Philip Guo: Good. That was a good question, so the questions about the role of teaching
because the book was all about research so I purposely made it very focused on, cause
like you can talk about all sorts of different things, you can talk about, like, what I do
in terms of teaching or extracurricular or, you know, other stuff but I doubt that anyone
cares about my personal life, right? People care about the process of doing the PhD and
getting research out. So, the question of teaching, so in our department we had 2 classes
that you had, you had the TA, 2 classes, minimal because I was on fellowship, I could do what
it was called full TA ship so I could only do a half TA ship which means they pay you
half but you actually do all the work. [Laughter]
>>Philip Guo: So it was actually the worst deal ever. So I ended up TA-ing 3 classes
and, for me, teaching was not a huge, the primary teaching for classes was not a huge
part of my experience. It didn't take up a ton of time and it wasn't a huge, it wasn't
a huge burden or a huge bonus cause, this varies by department but in our department
at least the teaching assistants, we didn't really teach like we just kind of maintained
the course and held in office hours and graded the work and so there wasn't really a big
teaching but there are some people who actually, if they actually like teaching more they would
try to teach a summer course. In that case they would actually get credit for teaching
2 courses and they would actually lead the course and give lectures. So, for people who
like teaching that was a probably better way to go. That said, though, the teaching experience
I had throughout grad school which were more gratifying, were the things I just did on
my own. So I have two experiences on that, one is I went to, um, I did some python tutoring
at a local, at a local startup that they contacted me cause, you know, I'd kind of been writing
about python in my blog a few years ago and then they contacted me and they said that
some other scientist and non technical people wanted to learn some basic python just so
they could just get a feel for the programming. So I went there for about 10 weeks, once a
week, and I would just write on the whiteboard and do introductory python stuff. And the
other thing, which is even more cool, is that just last year this guy contacted me, he was
a local entrepreneur here and he contacted me and he said that he was looking for someone
to teach him programming because he wanted to learn web programming and he's been at
this business for decades. He's a very successful business person but since he's in the tech
world he wanted to actually learn to code. So I thought that was really admirable and
really cool. So I actually ended up spending about 6 to 9 months with him. We did Skype,
mostly just screen share on Skype, peer programming for about 6, I have an article on my website
about this experience, it was an awesome experience. And he basically went from not knowing anything
to being able to build a non trivial, Facebook, social type of application. It was really
cool seeing him build everything and my role was basically as a tutor, so, you know, he
would just do everything, I would just send him documentation of like, oh look into this
library, and then he would struggle with it and then we would debug together and get him
over the hump, so that one on one was really, really spectacular. But, in terms of institutional
teaching, that wasn't a big part of my experience. That was a super long answer.
[Laughs] >>Philip Guo: Bianca can just field the questions,
yeah, you can. Microphone? [Pause]
[Inaudible]
>>Philip Guo: So, there seems to be what time?
[Inaudible] >>Philip Guo: Uh-huh.
[Inaudible]
>>Philip Guo: Good, good. Oh yes, yes, yes, where is this one? Yes, perfect.
[Laughter] >>Philip Guo: Thank you. Yes, knowing when
to quit. This does not mean quitting the PhD program
[Laughter] >>Philip Guo: but quitting individual projects.
Quitting the PhD program is also valid. Yeah, so what am I, so my first undergrad research
supervisor when I was an undergrad, when I visited, when I visited MIT later on in grad
school looking for a place, one of the things he told me was that knowing when to quit was
one of the most important things he learned in grad school cause he also had a very similar
experience as I did. He spent his first 3 years working on stuff that like totally didn't
work out and he just made a switch around in the middle and then switch and his last
3 years he did something else that was way more promising and now he's a full professor
at MIT. And somebody else, other people have this story too, so I don't know whether there's
like a great answer to that. It's, I think one thing is you can't quit too early for
two reasons, one is that you kind of, it's kind of like there's a losing face type thing
especially if you're a Jr student and you try something for a few months and you're
like, "Oh, I'm just gonna quit" then it doesn't look very good to, you know, your senior colleagues.
And, also, quitting too early may be bad because you might just have not accumulated the skills
to get over that hump. Um, that's a very hard question. I think, I think maybe it, maybe
part of it is just if you can find something else that's better, I guess, that's like,
that's also very precarious because I basically quit stuff without having the next thing that
I wanted to do but I just got very lucky in that. So I don't really, does anyone else
have a better answer? This is just outsourcing. [Laughter]
[Pause] >>Philip Guo: Better answer for knowing when
to quit. I know it's very important I just have no idea. Let me think, so I think that
like mid way is a good point. So, this is a somewhat good answer,
[Laughter] >>Philip Guo: it's like when do you know what
the middle is when you're serving through, right? So, I think if you're averaging, in
the US, if you're averaging a 6 year program then by your 3rd year, like switching advisors
or switching groups, I think that usually by your 3rd year you should really figure
out if you wanna stay with your defaults, you know, your, like that, stay with your
defaults or if the defaults are bad you wanna switch. But, it's very hard to switch later
on I would say. So, what year are you now?
[Inaudible]
>>Philip Guo: Okay. So this is exactly the time.
[Laughter] >>Philip Guo: Yeah, exactly the time you should
be thinking about that, yeah. Yeah, so I think that, you can't be quitting too late or too
early. And questions, or, here, yeah.
>>male #1: So as an undergrad [inaudible]
>>Philip Guo: Good, good. So, um, to repeat the question, So there's this thing called
like the PhD club which is the very simple fact that there is way too many more PhDs
being produced than academic positions, than positions as professors so, like we were talking
about offline, for every, every professor's training way more students than they can be
employed as professors, right? So, one cynical way to look at this is it's as a pyramid scheme.
A less cynical look at it and especially were lucky to be in more applied fields, is that
there are great industry options such as here and other places. And I think we had another
question offline about intentions of doing the PhD, right? Like, why I got into it. So,
I'll just tell my own story which is basically the prologue of my, the prologue of my book.
Just, this thing right here. So I decided to do my PhD because I had, I did 3 internships
during my undergrad at not very good companies and, like, it was just a terrible experience,
Like, I just happened that the places I worked at were, you know, you're very Dilbert like
cubicle farms and I could see that. So, on one hand, like I was the only intern at these
places so I was being treated basically as a junior full time employee which is cool
because it wasn't, this is a very different experience, you guys are very, you know, there's
a great intern program and there's, people really take care to feed and clothe interns,
everyone's wearing Google shirts. [Laughter]
>>Philip Guo: Yeah, people take good care to feed and clothe interns and they take really
good care of you here. But, when I interned back as an undergrad I just went to places
that were not Google or those sorts of top places and on one hand I got to learn a lot
about what the, you know, what the quote unquote world is like. I was basically doing all the
stuff that everyone else was doing. On the other hand I could see that, you know, people
who are right out of college working there, people who were 5 years out, people who were
10 years out, 15 years out, 20 years out, like, I did not really see myself in that
kind of life and they didn't seem super happy with their lives, they didn't seem super happy
with their jobs and it was just not a very good work environment. And one of my father's,
one of my dad's friends worked at one of these same companies, he was in another department,
and he would come over to my cubicle sometimes and offer, you know, stately unsolicited,
everything I say is on my website so this is all public, and he would just be like,
"You should learn to cover up your mon-" we have these giant CRT monitors so he said,
"You should learn to like get these manila folders" like he did, "And just cover up the
sides of the monitor" so he could say that he was blocking out glare but that was so
you could like surf the web so that other people walking by can't see what you're doing.
And then, you know, his whole thing was laying low and just making sure not to cause a ruckus.
So, I thought if that was the future of my career than I was, it was probably not great.
So there was a big negative motivation because that, if I was graduating with a computer
science or engineering degree and those would be the places I got exposed to that I was
working at, I didn't really want that. At the same time, the positive experience in
school, I kind of like the idea of doing research even though I didn't really know what it really
meant as an undergrad and I felt that, um, I kind of liked the idea of doing research
and those kinds of things. So I went into PhD with that, with that mindset. So, it's
probably not the most noble intention for doing a PhD. That's what like my top thing
was, but I think, I'm very glad that I did it and over the last, the past 6 years I've
learned a ton from my experiences. So it's probably not a really typical story of why
I got a PhD but that's just my own story. Um, and I guess there's a PhD glut issue but
I'll address that later. There's a person in the back.
[Inaudible]
>>Philip Guo: Yeah, so I heard great things about APM program, I don't know about personally.
This actually means a lot to me because, what I mean by this is that, I guess with the exception
of the, you know, the full time people who have been here for awhile and they're more
veterans, like, we are all here, like, we're all like below the bottom of the ladder. Like,
we're like beneath the bottom, right? Like, grad students and people who were just recently
graduated, like, we're the bottom of the bottom. But the thing is that you're all in your PhD,
you're all motivated to be here because you have a lot of creative energy, great ideas
and you wanna be pursuing what you love to do but there's this huge disconnect between,
you know, as young people, like as ourselves as young people we really wanna do something
creative and make a contribution and the fact that we are at the bottom of any hierarchy
that, either an academia or an industry or anything, we are always at the bottom. So,
how do you, how do you figure out how to push forward your own agenda and what you wanna
do when you're at the very bottom? And I think this is, I could speak the whole hour about
this but I'll summarize. This, I guess this leading doesn't necessarily mean leadership
of projects or people and such. It really means leading your own agenda and I think
the high little bit here is that you need to get someone who is powerful and influential,
interested enough in your work that, have you all seen the movie Inception? My favorite
movie of all time, sorry. So, basically you need to get them to think that it's their
great ideas and you are doing it for them. [Laughter
>>Philip Guo: That's one kind of way to put it. That, if you can get, and this also the,
this also goes with the selling thing, that one of those successful things that I was
able to do in my PhD was to get people, certain people who are very influential to be excited
about what I wanted to work on because they wanted to work on it as well and then they
were able to publicize it for me and give me great feedback and everything. So I think
that leading from below, a big part of it is finding people who are more influential
and are more senior and figuring out how you can be useful to them while still pushing
forward, sort of, what you want. It's obviously not perfect but very few of us are gonna be
able to say this is what I wanna do, mic is out, okay. We'll get one from the corner.
[Inaudible]
>>Philip Guo: Okay that's a good question, the questions about advisors being overbooked
with too many students. I kind of punted on that route because I actually just went off
and did my own thing which is good and bad. So I was able to be a little bit more independent
because I was self funded by fellowships and, also, even the first 3 years when I was working
on my advisors project with other students, I wasn't getting a ton of personal attention
either. So I think it, I think it really depends, like, some students thrive very well when
you have good advising and some students are better off when they're left alone and such.
So, that, if that's a concern to you then maybe one thing is to be proactive about seeking
your advisors time and like setting up regular meetings and stuff. So, that's the easiest
thing to do, I think. Next.
>>female #1: How do you actually feel about [inaudible]?
>>Philip Guo: Yeah sure, definitely I'll try to repeat all the questions. So, the question
is how do I feel about getting my PhD at Stanford particular versus other universities? I was
pretty agnostic to the actual department and such. I really like, I mean, I definitely
like, I came here because I, like, there was a great school and great everything and I
didn't really over plan things but one of the side effects of being here the last 6
years is I learned so much about what's going on in the industry and start ups and just
the Silicon Valley entrepreneurial culture and even though, myself, I'm not interested
in being an entrepreneur so much, I have friends who are and such and I really like that Silicon
Valley environment. So, I think one of the advantages of being at a school like Stanford
or Berkeley in the Bay area is the, kind of; you get this closer connection with the industry.
I think that was a big advantage. There.
[Inaudible]
>>Philip Guo: Great, so the question is theory versus system. So, yeah, I spent most of my
10,000 hours in grad school programming hunched over, right? And my friends in theory spent
a lot of hours scratching their heads, thinking about problems, writing on the white board,
you know, getting very frustrated about things. So it's a different type of grind. I think,
I don't know how my experiences would have been in theory cause I haven't really done
that but did you have any ideas about what things might be different?
[Inaudible]
>>Philip Guo: Okay. I think, this is my own like unfounded opinions but I have a feeling
that in theory there is more of a variance and, you know, it's not as linear in the sense
of somebody can make a really brilliant contribution their first year of grad school cause they,
you know, they've been doing theoretical math for many years and they come in and they can
publish one great paper right at the beginning cause there's a great insight. They have to
work alot at it too, it's not like it comes instantly. But I think in a systems oriented
field I think it's practically impossible for a first year student to be like, "I built
this thing and I evaluated it and I got it published at a top conference all in one year,
my first year." So I think the non linearity may be a lot bigger in theory than in systems.
If you're actually having to hack on stuff for your PhD, it's just much harder to get
stuff out there real early but I don't know otherwise. Let's do one in this corner.
[Inaudible]
>>Philip Guo: Okay, so the question is about like work being more applied and more, you
know, industrial rather than being more academic and such. Were you think like commercializing
stuff like building a company out of things? [Inaudible]
>>Philip Guo: That's right. [Inaudible]
>>Philip Guo: There's no proprietary issues it's just that like, it's just not what's
valued. [Inaudible]
>>Philip Guo: So there's always like lower tier things people are publishing on. I mean,
that's what I end up doing but still there's issues.
[Inaudible]
>>Philip Guo: I mean just conferences and journals that are easier to get in then the
top things. I mean, that's one compromise you can make is submitting less [inaudible].
It's ultimately up to your advisor. If your advisor is like you must publish a paper at
a top conference or I'll not let you graduate then there's no choice but I think compromises
have to be met somehow. Let's do one in the back in the green. Yeah.
[Inaudible]
>>Philip Guo: Okay, that's a really good question. So it's now that I have my PhD, what, how
am I finding a leverage, I guess what I learned in my PhD? Without getting into too much specifics
about what I'm doing here because I want this talk to be public, we can talk afterwards
about Google internal stuff, but basically as a high level summary, I, I think I got
very lucky that I'm here doing a certain type of work that I really wanna be doing and it
would have been very hard for me, it would have been hard for me to be doing this without
having got my PhD. Not because I'm using my PhD to assist in my work, that's just not
the case at all, it's cause I actually, just a lot of these lessons I learned about basically
trying to get myself in a position to do the work I like and to be leading from below and
to be getting influential people interested in my projects. To be making pitches to people,
like, all this entrepreneurial stuff I learned that I think it'd be very useful in industry
and other things. But, I'd be happy to talk offline about specifics. Well, right next
to him. Yeah. [Inaudible]
>>Philip Guo: That's a great question. The questions about reforming the PhD process.
I guess is there a need to reform how PhD programs work? And I think people looked into
this in the UK, for example. In Europe there's proposals from more streamlined PhD programs,
there's proposals for different types of funding models and things, I haven't really thought
about those issues myself. One of the main reasons is practical, is I'm not in a position
to do this sort of stuff. I'm not, you know, a leader in research who goes to summits to
talk, to debate with university presidents and talk to leaders and so I haven't given
these things thoughts myself. My, my, good segue here, my whole thing throughout my PhD
was that there was a game there that we all have to play and we cannot change the game
because we're, again, at the sub bottom of the rungs ladder. So the only way is to quit
the game and just don't play that all together, that's good cause no one's forcing you to,
or if you want to get a PhD and start a career doing something related then you need to be
playing the game. And this is not just for people who are in PhD. People who are assistant
professors have to play the game to get tenure as well. And the hope is once you get tenure,
if you're well connected enough you can, when you're in your 40s and stuff you can try to
change the game with reform and such but that's totally outside my scope. Good question though.
Oh, yeah. Oh, this question. [Inaudible]
>>Philip Guo: Yeah, you can read this offline also. Yeah, I'll repeat the question. Can
you get a mic or, we need mics, it's weird. You can have my mic; I guess I'll just stand
up here. Okay, sorry about that. I just feel awkward up here. So, yeah, that's a hard question
of, this is the paying the dues thing, so I wrote here not paying your dues but pay
some dues and I think that I paid my dues for a little bit too long. I think that there
is a point that it's very hard, especially in a very applied systems oriented field,
that it is, I have never, I've seen, very rarely have I seen first year students say,
"I have a great project idea" you know, a lot of people go into PhD thinking this, including
myself, "I have this great idea for a project and it's so cool and so innovative and I'm
just gonna do it. I don't care what anyone else thinks, I'm just gonna do it." And the
thing you find out really fast is that there is often a large disconnect between what you
find is interesting an innovative. So, academic research is a small subset of what is interesting
and cool and innovative, right? [Laughter]
>>Philip Guo: So there's this large set of stuff that's cool and innovative and there's
a smaller set that's some kind of research thing. So Google does a lot of research that's
not academic, right, so there's a smaller set that some kind of research cutting edge
stuff. There smaller set that is considered quasi academic research. So, some people might
be publishing off that. There's a smaller set that's considered respectable academic
research that is being published at the top venues. That is, the people that determine
that are the people who are senior in the field right now. And then of that there's
an even smaller set that you are capable of doing due to your resources or due to your
lab, due to your advisor, due to your own skills. So, I think it's very hard for someone
coming in to think that they are actually in that subset. And it's great to try to do
that yourself. So I actually, my first year I tried to do some of that stuff myself but
I quickly found out that that was probably a worse experience than having to just do,
having to just be on a, on a grindy project that didn't go anywhere. The good thing about
being on a, you know, basically everybody who I've talked to who's gotten through their
PhD successfully, everybody has burned that first 2 years. Like, if you're doing a six
year program, your first two years will be probably pretty useless. Does anyone sort
of agree with that? Or are you vehemently--.Yeah, people who are very successful in their PhD,
their first two years, everyone's just like yea, things just didn't work out. A lot of
it is the learning through doing thing, too. It's very, the most optimistic view is that
it's, you can't really get good unless you struggle on things. And in my first two years,
even though the projects didn't end up contributing to my dissertation, the, I guess, the sort
of immediate benefit was that I was, I was basically hacking a lot of low level c programs.
I was doing a lot of low level c programming and it was really gross, hard to debug and
stuff but because I had those thousands of hours of practice doing that, later on in
my, in my PhD I was able to, I was able to approach projects that other people were not
even thinking about doing because they didn't have these sorts of skills. So, you can view
it as skill building. But, in general, it's a very hard question. I think that you should
definitely not do this for like four or five years and hope that in one year you're gonna
finish something. So I think two years is fine because if you don't burn on it now you're
gonna burn on something else. I think we had one in that corner. Yeah, back there.
[Inaudible]
>>Philip Guo: Good. Good question, so the question is if I could go back in time five
years would have just gone to industry right away rather than doing PhD? And I personally
would not have because I wouldn't have gotten this book out of it.
[Laughter] >>Philip Guo: I personally would not have.
And also would not have gotten my current position out of it. So, without going into
details, basically I got my current job or particular job role, because of a side project
I did during my PhD. So the side project thing is another thing that's not on here. I didn't
talk about this in my book because it's not germane to the PhD experience. The PhD is
a great time for, I'm just going off a rant here, the PhD is a great time for, this kind
of goes with your question of what's the purpose of doing a PhD? Especially in a very applied
field, I was always getting reader responses from friends who, from some friends who were
interested in doing PhD and he was saying how now that he's been in industry for many
years, for five or six years and he said that he was, he sublet an apartment from a friend
in Boston who was a PhD student and those few months that he was subletting he just
had a ton of fun just being able to talk to people who were very intellectually curious
and smart and motivated and, you know, and he said that in industry a lot of times, you
know, everyone has a lot of stuff to do and they have priorities and such and it's just
very hard to get smart, motivated people together with some free time and lack of responsibility
to do cool stuff. So, the great thing about PhD that they're paying you almost nothing,
so you just don't feel bad about taking two weeks off to hack on something or
[Laughter] >>Philip Guo: Or, right? It's like, cause
things, there's deadlines where you have to submit papers to do stuff, but there are cycles,
right? Sometimes you're just not feeling it for a few weeks, so, you're not feeling it
for a few weeks and you just don't do anything and, or you just team up with your buddies
and you can hack on some, so like, around 2007, for example, we had, this is when Facebook
apps were really starting to come about and quite a few, you know, masters and PhDs at
Stanford, they were doing their research and they just spent their time at night just hacking
in the building and some of these people, you know, ended up making some money off of
their Facebook apps, or their iPhone apps and most side projects you do probably won't
turn anything commercially viable. But, it's just a cool time for serendipity to occur.
That's one of the reasons why I really enjoyed it and I think basically, I think the calculation
everyone has to make or the rationalization, I guess the rationalization afterwards is
like, "Did you gain more than you lost?" Right? So there are certain things you gain and certain
things you lost and you felt like you could gain more than you lost and it's a good call.
But maybe it's personal to everyone. It's really hard to make a generalization. Very
back. [Inaudible]
>>Philip Guo: That's a really easy choice cause I couldn't get any.
[Laughter] >>Philip Guo: I, basically, last summer, so
around this time I was here, around the time that I was interning here and I went to some
conferences to give talks and stuff and that was when I, the switch finally flipped. Like,
I was veering away from academia for a few years mostly because of my publication record
wasn't publishing in top tier places, I knew how competitive the job market was and I wasn't
willing to, you know, try to work on post-doc for a few years and go with that. But, last
summer I decided that the switch was totally flipped and I just decided I'm not gonna apply
for academic jobs and that was such a big sense of relief and, also, it was such a big
boost to my own creativity in the sense of, one of the advantages of academia that people
tout is that place, the closest place you can approximate to having creative freedom
or whatever. But, for me the decision to quit academia last year actually opened up way
more creative endeavors for myself in two ways. One is that I was able to look for jobs
and other non traditional roles and the other way is that actually got me to do my final
project which I actually visited Harvard to do, they actually opened up the community
for me to do a really cool project because I didn't have to worry about, "Oh, I need
to publish in top conferences" or whatever. I actually just met a professor at a conference
last summer and we just started chatting about research and I was just not afraid at all
because, like, before if I was like, "Oh I'm gonna apply for academic jobs I need to kiss
up to professors" and "what did I say?" and I got really nervous and everything. But I
was pretty much the only person in the conference who was just like I'll just say whatever and
do whatever cause everybody else, especially senior grad students and post grad students,
like everyone was on their top game, right? Shaking hands, pitching their research
[Laughter] >>Philip Guo: talking about let's get a grant
together and I was just going around just doing whatever and that just worked out really
well for me. So I actually did not do an academic job hunt at all. Here.
[Inaudible]
>>Philip Guo: That's a really good question. So the question was that I was on fellowship
for five of the six years of my PhD and most people in our department were funded by their
advisors grants as an RA and they taught sometimes as a combination. That was, that, that, I
thought about that a lot and even though I was funded by a fellowship and if you read
in my memoir, the first three years I acted as if I wasn't funded on fellowship. I basically,
it wasn't like I come here on fellowship and I can do whatever the hell I want because
in the end you still need three professors to sign off and let you graduate. So you still
need to be in good favors of professors and do sort of stuff they're interested in. So,
my first three years I acted as though I wasn't on my own funding, but the last two or three
years I had my own funding. I think it did make a big difference, obviously my own experience,
I was able to have more freedom and diverge and do my own thing. It's also a curse as
well because if you were doing your own thing you don't have institutional support so I
had to make a lot more proactive efforts to get my work published, I had to face a lot
more rejections, I had to basically just end to end, do a lot of my own stuff. I had to
do my own product definition, my own, I had to write all my own papers, you know, deal
with all the rejection notices myself; there was no team and people to do that. So, I think
that, again, it's hard to generalize, but the students, if you're optimizing for publishing
top rate papers and getting out with a good CV, the students I've seen do that the best
are the ones who partner with really good assistant professors with grant funding, they
can be fellowship also. Basically you partner with someone, this is, sort of this, but if
you partner with someone who is like dead set on publishing because they need to publish
for their lives because they are assistant professor and they need to be publishing top
papers, top notch papers, if you partner with them and you get along well with them then
you just, there's this great synergy that happens. And the students I've seen do really
well in that, that is if you want to optimize for publishing papers and graduate with a
good CV. I did not really have that myself so that's a big part of the reason why I didn't
go for an academic job because I didn't have a CV that looked good for academia. I think
we have a bunch of questions, back there. [Inaudible]
>>Philip Guo: That's a good question. So do I ever think about quitting? So, I never thought
seriously about quitting. There was, this one down here, I never thought seriously about
quitting but I think a part of it was because of funding. So I'm like, I have this five
years of money then I might as well use it and just
[Laughter] >>Philip Guo: Might as well use it and there
wasn't, because, again, because I was used to, my perception of jobs in industry was
not a great perception because I didn't intern at great places. So I'm like, if I quit I'll
just go get a job in industry, I was tainted by that bad perception I had at the beginning
so I'm like, I don't look any different on paper now so I don't really see, the thing
is being in Silicon Valley I think if you get really good job, I mean, I had friends
who definitely left early and they were able to get very good jobs. So, yeah, it really
depends on everybody. But I personally didn't. Another, back there.
[Inaudible]
>>Philip Guo: Good, so a question about the actual book writing process. So I basically
decided about six months ago to do this and I just started collecting notes together from
old emails and just like my research notebook, like text files and notes that I kept and
various rejection paper notices that I could quote from. And, so, basically I just started
compiling together, I did this on the back burner, right? So as I was writing my dissertation,
in my spare time I would start refining my notes. So this is the way I often work, I
can only do one thing at once, but I have one background thread that I can do something
else. [Laughter]
>>Philip Guo: So while I'm doing my dissertation writing for three months in the back I was
molding this outline and then once my defense was done and I passed I could free up my time
to doing this writing full time. So it only took a few months. But, and the content is
actually not any, this book was all original content. It wasn't taken from other writings
I had which, you know, cause I didn't wanna be redundant and other stuff. Here, I'll take
yours offline. [Laughs]
[Inaudible]
>>Philip Guo: Right, right. I think it's hard to know, I mean, to know what the game is
in your department, I think just talking to senior students is the best way. Just talking
to senior students in your group and other groups and stuff there is usually some department,
nothing is explicit, right, there's not like, I don't think anyone's advisors like, "You
must publish these two papers at this conference and I'll let you graduate." It's all kind
of fuzzy and I think that from professors perspectives, they don't really like to talk
about it cause they don't wanna make the PhD out to be some, you know, checkbox thing where
you, I checked my paper off, I checked another paper off. Everybody has their own, there's
professors that don't want their students published at all cause they just want them
to write one long 200 page thing that's amazing and detailed and don't need to worry about
publishing. So, I think you just have to figure out by, you know, the best thing is if you
can talk to your advisor about it, that's the best obviously, but that's, may be hard
to do. But I think it's just talking to people around you. Okay, great, thank you. Yeah,
here. [Inaudible]
>>Philip Guo: That's a great question. So the question's about how do you balance helping
other people versus, this one here, how do you balance helping other people's projects
versus your own? I think one very rough rule of thumb is that if you're in, let's assume
you're doing a six year US program, just assuming six years, it, doing your first three years
I think taking every opportunity you can to help other people is probably the best because
that's, helping senior students and post-docs, that's where you're gonna learn a lot from
and that's your most likely chance of getting a paper published and even if those papers
don't go towards your thesis, it's just the idea of being able to practice end to end
of hacking on systems, writing papers, dealing with rejections, submitting papers cause there's
just a whole art about it. And if you can partner with someone by helping them then
they're obviously very motivated cause it's their project and as you tagging along you
get some credit and you learn a lot. I would say that in your last few years, that you
should be a lot more hesitant on it, because, you know, you wanna really get your foot out
the door and, I think it's more justified for you to be more selfish later on, too,
cause you need to graduate yourself. Alright. [Inaudible]
>>Philip Guo: Right, so the question is about the paper writing process. This, again, varies
a lot by field so in computer science and typical computer science fields you're trying
to be publishing in these kind of top tier conferences and basically that involves writing
a pretty dense 10 page, or so, paper and you submit it, there's a deadline for every conference,
you submit it you get feedback within about three months or so. But the feedback is really,
I mean, you don't get to revise it often times, it's you get it and it's either a yes or no.
If it's a no there's another conference coming up that you can resubmit to. So, usually for
one paper you can probably submit it about three times a year, two to three times a year
and, you know, the trick with that is you really have to balance out, once you submit
one paper you have to be either working another project or essentially a paper, you really
have to time it. So you can't just single a task and say, "I wanna get this paper in
no matter what" and not do anything else. You can if your advisors like, "You will get
this paper in somewhere and you'll graduate" and just focus on one thing but there's definitely
a strategy involved in, you know, in getting feedback and such. Let's do one all the way
in the back. [Inaudible]
>>Philip Guo: How do I think an advisor would react if you just came in wanting to do cool
stuff? I think if you're self funded than they'd be okay with it because, you know,
they're not paying for you. So if you're self funded you could do that. I mean, I wanted
to sort of do that my first few years but I kind of had an eye, not really on, not really
on going to academia as much but as in I need to play the game to graduate. So I think that
if you have your own money it's good, if you don't then it may be harder to convince them.
That just depends on the person, if you're very charismatic and very good at convincing
people then that's great. I mean the longer term you wanna think about is how you're gonna
get a PhD out of it or if you don't want to, if you just wanna hack on cool stuff for two,
three years and then get funded and drop out, that's legit as well. So it just depends,
I think doing cool stuff especially when you come in the beginning, that doesn't have a
very high probability of getting you published and getting a PhD but if your goal is to just
do cool stuff, not necessarily get a degree and just be funded for a few years and do
cool stuff, that's a good route to go. Yeah. [Inaudible]
>>Philip Guo: Yeah, so one of the, yeah, so I never had that problem cause I didn't have
that many choices, but if you have a lot of ideas that you could be working on maybe one
way to optimize is, you know, what's the one that is most likely to produce something tangible
first? Whether it's a publication or it's a, you know, some kind of milestone and I
think this kind of output trumps input thing is kind of relevant is that one cool way to
look at this is your PhD is all about your output, right? And output for most things
is about papers or your thesis or whatever and so if you have 5 ideas and one of them
is most likely to result in something then I would maybe go with that one because if
you get some output then you can get some feedback and get some, you know, you can have
a launching form of something else because it's all too easy to work on some half baked
things that don't pan out. So, okay. [Inaudible]
>>Philip Guo: Yes, it's very Meta, right? Yeah, so I think, I gave about 25 or so talks
throughout my PhD. So I gave like one talk every few months and this is one of the pieces
of advice that, like, professors always tell you about that like no one really gets it
when they say it, they're like, "Oh, you should talk while you work." Giving talks has been
one of the most useful things for me, at all stages, so if you're just forming a project
or starting out, if you can just give a very informal talk to your group mates at lab meeting
or, you know, something within your school of students, they can give you like great
early feedback. It's basically just a way of getting early feedback on what makes sense
and what doesn't make sense on a very superficial level. And as you develop your project and
you're about to submit a paper, like giving a talk right before you submit a paper is
really useful because people are gonna come up with common criticisms like, "Oh, why didn't
you think of this?" or "what do you do about this?" and those common criticisms are what
you need to address in like the very beginning of your paper, right? Cause what, the talks
are really about someone getting a superficial view of your research and paper reviewers
get a very superficial view of your research because they have a stack of 30 papers to
review. If they don't like your paper within the first 2 minutes, they're gonna dismiss
it and find ways to reject it rather than accept it. So you need to make an amazing
first impression, right, you need to make an amazing first impression on the first page
and giving a talk right when you're about to submit a paper is a great way to ferret
out those really easy-- it's just like useability testing, right, very simple form of useability
testing. You're kind of testing useability of your presentation. And then, obviously,
giving the more, it's funny cause giving the more formal conference talks to, you know,
quote unquote important people, those are actually much less useful because you're at
a conference and everyone is on their laptop and no one really cares who you are and stuff
and you're just kind of giving it because you have a paper and you have to give a talk.
So, for me, the more informal talks are actually very useful. So I gave a tech talk here last,
last, uh, two years ago or whatever and somebody, amazingly, people came to my tech talk who
I didn't invite. Like, there were actual people who were actually interested who weren't my
friends and there was some guy in the back who was like pestering me with questions and
I thought he was kind of being annoying but then he ended up, like afterwards, he's like,
"I'm really interested in this work. Why don't you come intern for me this summer and we'll
just let you work on this, your own open source project all summer cause I'm just interested
in these topics." And I accepted my internship and that's what I spent last summer doing.
I was, basically I spent all my time working on my open source, my open source research
project here at Google under his, you know, protection because he was just interested
in having me around. So, that's the unexpected serendipity of giving talks. Another one was
that after a talk, because after you give a talk people have their contacts between
them. So somebody, another grad student emailed me and said, "Oh, did you see this blog post
by blah, blah, blah?" A related topic, then I saw the blog post and then I emailed this
guy, a few months later, he was at Berkeley, I emailed this guy a few months later saying,
"Oh, I'm doing something related" and we chatted and that was basically the foundation of two
of the five projects I did for my thesis came out of talking with that guy which would have
never happened if I didn't give another talk and another student hadn't sent me a link
to his blog. So weird stuff happens when you give talks. Um, yes, here.
[Inaudible]
>>Philip Guo: That's right. So the question is how easy or hard is it to evaluate people's
motivations looking forward? I think, I think that you just get better at it every iteration
through. So, there's nothing perfect, so while I'm starting my career here, I think that
I'm better at evaluating the motivations of my superiors than I was my first year in grad
school. That's another PhD learning thing and you can learn this in industry, you learn
this everywhere, right? If you work with enough people, you just get more repetitious in doing
it, but, I don't think there's an easy answer. But, obviously my thing was written in retrospect,
so. [Laughs]
>>Philip Guo: Let's get one last question. And people can come up to me offline, I'll
be happy to chatter. Um, the red shirt. [Inaudible]
>>Philip Guo: Yeah, so if I had interned at Google during my undergrad I would probably
be at Google without my PhD. Yeah, if I had, back then in those years, Google was, obviously,
this was around 2003, 2004, so Google was very up and coming, Microsoft was a big place
where, you know, Microsoft was the Google of internships back then, like that was the
top internship program. My friends who went to Microsoft back then mostly went full time
there. So if I actually would have had an amazing intern experience I probably would
not have done my PhD. I would not be giving this talk here. Cool. Thanks a lot and I'll
be around afterwards for questions. So, thanks. [Applause]