I explain Artificial Intelligence terms and news to non-experts.
The best videos on the Internet archived and shared on HackerNoon.
The best podcasts on the Internet archived and shared on HackerNoon.
यदि आप पीएचडी करने पर विचार कर रहे हैं। मशीन लर्निंग में, या केवल "उचित" पृष्ठभूमि के बिना क्षेत्र में उतरना", यहां कुछ महत्वपूर्ण अंतर्दृष्टि हैं जिन्हें आपको निर्णय लेने से पहले जानना चाहिए।
एक पीएच.डी. मशीन लर्निंग में क्षेत्र में प्रवेश करने और विशिष्ट अनुसंधान लक्ष्यों को प्राप्त करने का एक शानदार तरीका हो सकता है।
हालांकि, अकादमिक बनाम उद्योग के पेशेवरों और विपक्षों पर विचार करना महत्वपूर्ण है। ओपन-सोर्स रिसर्च संगठनों में शामिल होने या साइड प्रोजेक्ट्स पर काम करने जैसे वैकल्पिक रास्ते वास्तव में कुछ व्यक्तियों के लिए स्नातक अध्ययन से ज्यादा उपयोगी साबित हो सकते हैं।
रेज़्यूमे को वैयक्तिकृत करना : जब मशीन लर्निंग में अपनी पहली नौकरी देने की बात आती है, तो आपके द्वारा किए गए मात्रात्मक प्रभावों को प्रदर्शित करना महत्वपूर्ण होता है। यह किसी सेवा की गति बढ़ा सकता है, किसी मॉडल के प्रदर्शन में सुधार कर सकता है, या किसी नियोक्ता के लिए राजस्व उत्पन्न कर सकता है। इन उपलब्धियों को उजागर करने के लिए अपने रिज्यूमे को वैयक्तिकृत करना महत्वपूर्ण है।
ब्रांड और ऑनलाइन उपस्थिति की शक्ति जब भर्ती की बात आती है तो एक मजबूत व्यक्तिगत ब्रांड और ऑनलाइन प्रोफ़ाइल के प्रभाव को कम मत समझो। अपने काम को साझा करें, सहयोग करें और अपने लाभ के लिए सोशल मीडिया का उपयोग करें।
कनेक्शन बनाएं : किसी भी क्षेत्र में नेटवर्किंग आवश्यक है, और मशीन लर्निंग कोई अपवाद नहीं है। पुराने स्कूल के सोशल नेटवर्किंग में गोता लगाएँ और उद्योग में संबंध बनाना शुरू करें। लिंक्डइन पर लोगों तक पहुंचें, नेटवर्किंग कार्यक्रमों में भाग लें, ऑनलाइन समुदायों में शामिल हों और प्रतिस्पर्धी बाजार में अलग दिखने के लिए अतिरिक्त प्रयास करें।
एक कहानी साझा करें : एक बार जब आप साक्षात्कार में शामिल हो जाते हैं, तो अपनी कहानी साझा करना महत्वपूर्ण होता है। रिक्रूटर्स ने आपका रिज्यूमे पढ़ लिया है, इसलिए उस प्रोजेक्ट के बारे में बात करें जिस पर आपने काम किया और जिन चुनौतियों को आपने पार किया। जो कुछ भी आपको बनाता है, आप!
इन जानकारियों को ब्रायन बर्न्स , पीएच.डी. द्वारा साझा किया गया था। वाशिंगटन विश्वविद्यालय में उम्मीदवार और एआई पब ट्विटर अकाउंट के संस्थापक। एआई में आने के तरीके के बारे में और भी अधिक जानकारी जानने के लिए, एक ट्विटर पेज विकसित करें, एक पॉडकास्ट होस्ट करें, इक्का-दुक्का साक्षात्कार करें, और एक बेहतर रिज्यूमे बनाएं, नीचे दिए गए पूर्ण पॉडकास्ट में ट्यून करें!
foreign
this is an interview with Brian Burns
PhD candidate at the University of
Washington and founder of the AI Pub
Twitter account now if over 60 000
followers Brian is trying to build a
talent referral business out of it Brian
also hosts a podcast where he explained
research papers with the authors in this
interview we go over all these different
facets the PHD I think a Twitter account
the podcast and more I hope you enjoy it
who are you and what's your background
sure
um my name is Brian I run this Twitter
account called aipob an Associated
podcast called Deep papers
um
what do I do
um with the with the Twitter account I
cover like high level
technical AI topics so it's like AI
research but not going like super deep
in the weeds but also for people who
have a technical background
um and then and then the podcast is like
interviewing uh AI scientists of like
recent like very interesting papers also
like on a technical but high level
um my background is a is uh before that
I'm I'm on leave from a machine learning
PhD within the stats department at uh at
UW I I did a year of a stats PhD uh
picked up a lot of like deep learning
stuff some computer vision stuff
um and started this Twitter account it
blew up and I'm kind of I'm kind of
running with it so that's uh that's my
background we'll definitely dive into
the podcast and the Twitter account
which is my main interest here but I
also want to take a small detour into
the PHD just because I'm myself doing
one and I know that a lot of people
listening are either doing one or
considering doing one so I'd love to
know why well first what was your your
reason in starting one
yeah I mean for me I feel like my
situation is kind of weird like for me I
wanted to break into machine learning uh
I didn't have I had a programming
background I've been programming since I
was a little kid or a teenager
um they had very very strong math
backgrounds so like growing up I wanted
to be a mathematician and like in
University it was kind of on on the
track to like do a math PhD
um I wanted to break in machine learning
like uh I thought that basically like
doing a statistics PhD like in a
department that had a really good
machine learning program would be a good
way to do that
um so that's like so I ended up applying
to a bunch of grad grad schools and I
got into UW
um so that's kind of like why I started
it
so to transition from mathematics into
something more applied
yeah I I just uh I I wanted to like
learn a bunch and like grad school is
like a really cool opportunity where you
know you get paid enough to pay rent and
buy food and you know you can use a lot
of your time not all of your time but a
lot of your time to to to learn um so
for me it was really useful
what would you say that it was maybe
more useful than directly going into
some industry roles like what's the
what's the difference for you or or why
did you choose the The Graduate Studies
path instead of going straight into the
industry yeah I mean honestly I think
that's a good question I I think in
hindsight
I think in hindsight I didn't understand
the breadth of opportunities that are
available
um I definitely think that having the
credential really helps even if even if
you do a fraction of a PhD like I have
I'm on leave I almost certainly am not
going to finish
um you know having that credential or
having you know master's degree or
something it does help but I do think
there are kind of a lot of ways to break
in and I
I definitely don't think that the PHD is
necessary at this point especially given
how quickly quickly things are moving
and how quickly kind of the dominant
paradigms of AI and machine learning are
changing um I I don't think it's I don't
think it's necessary so it definitely
like in hindsight I don't think it's
necessarily the optimal path and and I
think for for a lot of people it's
definitely not yeah and do you think
regarding the optimal path do you think
like of course the credential helps but
do you think that the four years put
into it well like minimum of four years
or something like this put into it is
worth the four years that you could have
put into
either working on your own thing or
working at a startup because let's say
we you won't be able to enter Google
right away if you don't have studies but
like do you think the time invested in
the PHD is worth the certification
I think it just really depends on what
your end goals are
um because I think I think there are a
lot of reasons people will pursue a PhD
um or kind of you won't want to do
something in machine learning I mean if
you're if your goal is just to like
learn and do novel research in machine
learning I think like if you can get
into a PhD program or somehow fund
yourself as an independent research
researcher uh or even like work with
these um
these open source Labs like like
stability I I think that's the best path
uh like if you just want to do research
like there's kind of not
there's not a huge societal slot for you
unless you either do a PhD or you you
join one of these open source research
Works
um I think for many of the other things
you want to do I don't think it's
optimal like I think I think if you want
to
just like get on the ground running and
learn how a lot of these modern machine
learning systems work and like you know
find good work in the area I think you
can you can learn a lot of this stuff
really quickly on your own um or even on
your job
um so I think for the vast majority of
situations I I wouldn't recommend it but
I think if you wanted if you if you're
really dead set on doing research and
you can get into a good PhD program
um I think it's probably the optimal
path
um
and then and then the other thing is
like again like the world is changing so
much like now there are these really
cool open source Labs like I have a lot
of friends who are like software
Engineers or machine learning Engineers
where they want to increase their
research shops so what they do is
instead of leaving for grad school they
just start working with like Carper AI
in their spare time
um so I think even if you're trying to
do research I think there are alternate
paths there
yeah definitely and I I don't know what
will happen with universities in general
but we will see and I I also think well
I assume that I assume it's obvious that
if you want to become a professor as
well it's pretty much necessary like if
if your goal is yet to teach it it
definitely helps and well it's not
necessary you can become like an online
Professor or something but
if you want to teach at at University
it's definitely required and to me that
well after speaking with many people I
think even doing research as you said
might be might not be the optimal path
the PHD might not be the table path you
may
be even better if you if you are able to
find some research position elsewhere
and the yeah the main reason to do a PhD
might be to become a professor and yep
the second one to to do research which
may not even be the the optimal path to
do that yeah I mean I think if you want
to do if you want to do research doing a
PhD is kind of the mainline way to do it
like I think there are alternate pads
but like whatever 80 90 percent of
people who are trying to pursue research
and who can get into some kind of PhD
program like I think that's probably the
safest bet and the most straightforward
way to do it that would be my guess
yeah it's also
but it's definitely worth doing if you
enjoy doing it like it's it's definitely
if you like to to control your own work
and pace and for example if you want to
take eight years to do the well I don't
know what's the maximum length but if
you want to do for example seven years
instead of four but work 20 hours per
week I most times you can well here in
Canada you can so like it's definitely a
good thing if you are more I don't know
that's a bit weird but I I can't see a
comparison between the PHD and
Entrepreneurship where you pretty much
decide on what you are doing and you
decide on your schedule and you work on
what you want to work so
that's that's also pretty cool and the
yeah the the flexibility that you have
is amazing to work on side projects just
like yours so I feel like it it may be a
good like
safe packet plan while trying other
stuff at the same time yeah I agree with
that I agree with that 100
and so you mentioned it but I assume you
most certainly will not go back to the
PHD and continue it
yeah I would say so
um I think
I think I've kind of explored this weird
alternate path uh of you know like
social media is just so cool like I just
have met all these different people in
social media over the last
six months or so seven months and
starting the account like I you know I
have this little recruiting business
that's going well
um I just there's so many kind of
opportunities that I feel like are here
um that I think at least at least for me
I think like I don't I don't necessarily
want to do another four years
yeah that makes sense yeah but it's very
very different because I I think for me
like I entered the PHD to kind of break
into machine learning and I think after
the first like six months to a year I
felt like I had learned enough to go out
on my own
um but but I think like I was not dead
set on like doing you know world-class
language modeling research or something
like that and I think I think for
someone else if if that's if that's your
aim then then staying is really the
right move
yeah and it's definitely something that
you are like in a PhD you are
specializing yourself in something
extremely specific and that's well that
that brings value but it may not be for
everyone yeah and it's also I mean this
is one of the things that like blew my
mind the most about uh machine learning
kind of the history machine learning is
um is almost the trend of
anti-specialization with time like most
Fields as they go on you know they get
more and more Niche it's like if you do
like a physics PhD or something you know
you specialize in like
theoretical Quark String Theory type B
or something and you do like seven years
on that whereas the really interesting
thing with with ML over the last like 10
years you know post post deep learning
and especially with the kind of like uh
you know Transformer sequence modeling
Paradigm is like this unification of
architectures right where it's like yeah
you know I I'm not an expert on this
stuff but it's like 10 years ago like
people were doing computer vision people
who were doing like NLP were doing like
very very different things
um and now like a lot of the methods are
actually becoming more and more similar
um so it's it's also kind of an
interesting time to like think about
specialization or like really deep
specialization
you know
yeah and I also I've read a great book
well listen to a great audiobook about
uh I don't remember the name exactly but
it's like why generalist Triumph or
something like this which was really
interesting and about how
generators are often the ones that that
succeeds even in in some very specific
fields for example I remember one of the
examples he gave was Roger Federer the
tennis player he they in in the book he
told he he wrote about his background
and said that like
um compared to other tennis professional
players Roger Federer where was playing
like
every other sports and some tennis until
he was yeah I would say random numbers
but until his things or something
compared to others that like at six they
already started only playing tennis and
nothing else whereas he was playing a
lot of different sports and developing
different skills and so I I don't know
why I'm biased towards the
generalization because I just like
touching a lot of stuff but I feel like
there's there's definitely some value to
to staying a bit more general or at
least
even if you are focused on one area
still getting out of your way to
understand the different things that are
happening or developing in other
Industries
oh yeah I mean yeah I kind of I see this
a little bit in recruiting as well like
I do Talent referrals to different AI
startups and I think like the candidates
that I find like get the most job offers
or like end up like progressing in their
careers the fastest they're definitely
not like total generalists like they
tend to be really good at one thing
right yeah
um you know they tend to be like really
really good at NLP and like have a long
history of projects and like you know
successful work history doing that kind
of thing but I think it's kind of like
you're they're not entirely t-shaped
it's like they're really good at one
thing but then for basically everything
else they also are like good enough or
or actually like even a little bit
beyond that so it's like they're really
good at NLP but then like they're also
like really fun to work with and you
know they have at least some social
media profile showing off their work and
like they Network every once in a while
you know so that I I I've noticed like
that that additional kind of marginal
stuff kind of across the skill
distribution like really helps people in
their careers yeah or even for example
just just for myself in my internships
and things like that if you are good at
something else it can also help just I I
remember that I I
we I wrote in my resume that I was doing
some climbing and like one of the
interviewers were was a big climber
climbing fan and we just talked about
this during the whole interview and I
ended up getting the job and we just
went climbing and it was
perfect like I I feel like a lot of
recruiting is based on
and not preferences but like just fine
it's just like finding a friend
basically you want to match with the
person you like the person can learn and
as long as he's he or she is motivate
motivated it's it should be fine so I so
you definitely try to
to find a good fit but you are the
recruiter expert so what what would you
say that it's most important when when
matching like it I don't know matching a
company with with an individual like or
what what would you say that companies
are looking for mostly is it strictly
the skills or the personality is even
more important what's the most important
thing oh man that's that's so hard
because I don't think there's any one
specific thing and I think there's also
a selection for a lot of different roles
right like I think what kids you had as
a research engineer is different from
what gets you out as a software engineer
research scientist so I do think it's I
think it varies a lot across roles
um but let me let me answer directly
um
so I think so there's there's kind of
like two levels of this like I think uh
you know first and foremost I think just
like having impressive technical
accomplishments on your resume is really
useful like this is just very standard
recruiting advice but it's just like
you've shipped a bunch of projects with
like very quantifiable impact yeah you
know n dollars a year for your business
for your employer or you know you sped
up sped up X service by X percent or you
know you improved X models performance
from you know A to B
um so I think that stuff matters first
and foremost I hate to say it but I
think second is like brand I think
recruiter like people who are hiring
CEOs or startups you know or careers
like the the name on the resume matters
a lot you know like if you're working at
whatever Google Facebook or like you
know a big startup uh stripe Airbnb or
something like um that it really greases
the wheels to to get you in the door
um you know whether whether or not
people like that the other thing to say
about the brand stuff is I think um I
think it's more in your control than you
think
um so it's like it's harder to you know
change whatever company like maybe you
can't work for a really prestigious
company but
um I think if you like if you end up
going on social media and you make a
profile for yourself and you like share
really cool work of yours you have like
a really slick website
um there's a way that that actually can
really like alter people's perceptions
of you and like show off stuff that
doesn't show in the resume so I think I
think second is brand like below the
technical stuff
um and then I think third I think third
the thing that really helps with
recruiting is kind of exactly what you
said it's actually a lot of this like
really like non-scalable social stuff
um it's really funny I think and again
I'm like not an expert with recruiting
I've been doing this for maybe six
months or something five months
um you know there's all this stuff like
there's job boards you know there's the
job application page and
one of the things I've learned is that
just like a lot of this stuff just gets
totally ignored
um there's just so much volume on the
internet that like these job boards or
these these job applications which get
totally filled up and often the the CEOs
of the hiring managers are so
overwhelmed that they just don't even
look at the uh at at the applications at
all like it's it's really surprising so
like the funny thing is kind of I think
the thing I've learned about recruiting
and like job stuff is like if you want
to get good at it you have to get good
at like all the old school stuff so just
like networking or like finding out who
the hiring manager is on LinkedIn and
like sending them a really nice cold
email about why you think you're great
for this position so that kind of thing
um so to recap One technical stuff
matters the most to Brand three like
non-scalable old school social stuff I
think referrals to like like having a
friend who's part of the company or
something like that
and I'd say like fourth that is outside
of your CV would be your online presence
as you mentioned just it's really
important to to be to be visible
basically it's it's not it's I mean I'm
sure you've experienced this like just
the kind of opportunities it opens up or
the people you mean it's totally totally
Bonkers totally bonkers
yeah it's really like if you guys are
not on either Twitter or LinkedIn just
go create an account right now and at
least share what you've done in the past
year or something just just be online a
little bit and if you are coding on
GitHub you can also put it to just
public and if you can put your code to
public and clean it a bit before but
yeah it's
really a game changer to to have an
online presence especially in this new
field that where we see for example I've
interviewed someone
um the VP of of research and development
at a startup called did and they he said
that for a research scientist role they
they received in the thousands of like
um potential candidates that were
qualified that's just crazy like they
cannot go through all of them so
definitely you need to stand out one way
or the other and as you said either by
reaching out and sending a personalized
letter or at least use chatgpt or
something to write it like more personal
but like not not spam them but at least
send something that seems personal is is
already a good step to to to get out of
the out of the lot out of the the bunch
so yep
so you mentioned that yeah we talked
about recruiting and talent referral
your what is I believe your current
business so or like main business model
maybe but what does this mean exactly
what what are you doing and how do you
do that how how can you
how are you matching people and finding
candidates and everything around that
yeah it's um it's it's actually quite
simple and I uh a lot of it I'm still
experimenting with it and I think there
are a lot of different things that can
be done but
um at least for now how it works is I
just have a link on my Twitter page and
people can click on the link and they'll
you know submit some information about
themselves their name you know what what
company they're working at like what
they're looking for in their next role
um and then you know typically like if
if they uh if they uh meet certain
parameters and I'm explicit about this
stuff it's like you know fortunately or
unfortunately uh almost all the startups
are referred to are in the US so it's
like if they're in the US and they have
like two years of experience
um and they submit that stuff I'll reach
out to them for for like a phone or a
zoom call just to get a better sense of
what they're looking for
um and then they're just kind of like a
passive Matchmaker so like uh at that
point maybe you'll really understand
what they're looking for and already
know a company that's that's a good
match
um or it's happened where it's like
WeChat I don't actually know anyone
who's a great fit for them but then like
three months later I meet startup
founder who's like looking for someone
who you know fit the profile
um so then you just put people in touch
via email
um and that's um that's how it works I
I've started to there's one company I've
been recruiting for that won't say much
about them
um but they're like really cool I'm like
a huge fan of them uh you probably
guessed who they are if you like follow
me on social media
um I've started to work with them a
little bit more closely So like um
started to like do more Outreach like
like actually doing outbound and you
know either reaching out to a couple
people who I follow on Twitter saying
like hey I think you'd be a really good
fit here
um or even like on LinkedIn
um
but that's actually this is the first
time I've ever done like any kind of
outbound Outreach
um so that's that's currently how how my
thing works
and you mentioned that this was mainly
thanks to your Twitter account so I
think we we should definitely dive into
the the Twitter which is called AI Pub
and for the people that are not familiar
with with it if you're in the field I I
think we can assume that you are
familiar with it just because of all the
reads that you've got in the recent
uh I don't know six seven month or
is it that much around there I think it
started in like September again
I think that you you've got something
like 8 million reads or even more now oh
I I actually haven't checked the numbers
in a long time I know I have 60 000
followers uh yeah something like that
yeah definitely I I would assume the
fastest growing Twitter account
especially in in artificial intelligence
and you already explained a bit what we
you were doing but could you go a bit
more in detail of what is your goal with
the satorical as well as
who is it for who should follow you and
and go check it out yeah I think um
again like all this is kind of an
experiment like here it's kind of
there's nothing really set in stone
um so I'm kind of playing with a lot of
different things
um I'll start by saying like who is my
target audience I think target audience
is like technical people who are in the
industry or like very adjacent to the
industry
um so like people are professional
software Engineers people with
professional machine learning Engineers
um who either like read machine learning
papers themselves or like kind of want
to be kept abreast of it all
um
you know uh so that my target audience
is like generally technical people
um and then what I'm kind of trying to
do is cover
technical AI topics at like a high level
so like research summaries but like not
going super deep in the weeds so that
people can understand a lot of the high
level Concepts
um without spending hours and hours
going into the nitty-gritty uh details
so there's just a lot of stuff that
interests me
um I think I think over the last couple
months it's mostly been
um
just like summaries of research papers
um for a while I did this thing that I
feel really bad about I've just been so
slammed with the recruiting I haven't
kept up with it but like for a while I
did I did like a weekly Twitter thread
that like summarized all all the things
that happened this week on AI Twitter
that must be crazy right now if you were
to do it no yeah I actually again I feel
terrible about this because I haven't
done this in a couple weeks because the
recruiting stuff um but yeah like I
don't even know how I would do that now
because it's like the Twitter thread
like you know you should probably do
that Twitter thread every two days or
something like that at this point are
you familiar with Alpha signal the yeah
so like he he posted I think it was
Tuesday like people said that summary of
what happened Tuesday and it was like
more more new things that what you were
sharing in a weekly thread yeah no it's
it's it is it's it's it's it's kind of I
I my emotions are rare mix it's like
very exciting very overwhelming a little
bit terrifying depending on how you
think about the future
um so I think I think all that stuff is
interesting I think one thing like I
kind of want to explore in the future
with the account is not just like like
research progress but also talking about
kind of like how AI or is affecting
Society or is affecting the economy or
is affecting the startup scene um
there's there's a podcast that's run by
a friend that I really really like that
kind of covers all this stuff super well
it's called the cognitive Revolution
um and I think like
my plate is kind of full right now but I
think if I get more bandwidth I think I
want to do more stuff like that because
it's just a personal interest of mine
like aside from just keeping up with the
research papers like I think this stuff
is just going to be so transformative
for society and like even just like
walking around or driving or doing
errands like I'm just thinking about
like what's the economy going to be look
like in three years or like how you know
how are gonna people going to have to
adapt you know what what should be
people be doing to like ride that wave
um so I think like that's also a
direction I kind of want to take the
account of you once I get more bandwidth
and I think my Powers would be really
interesting because I think I think
there's a thing that's on like
everyone's mind right now anyone who's
following gpt4 or any of this stuff like
they're thinking about this kind of
thing
yeah exactly and as you just said it but
I think that would be extremely relevant
to
technical people that usually don't
think about this stuff but they are
developing those models and working with
them and Etc but they
they are
for sure interested in what will happen
and how it affects Society but they
don't have really the time to
investigate or think about it so it will
definitely be a great transition for you
I think well just just because you are
you already are building the to my mind
the perfect audience that should be
um sensibilized with that so that that
would be definitely great to see I'd
love to see that yeah if I I'll just say
one thing I I think they're like all
these different like media niches even
like within machine learning like
there's kind of like machine learning
for like people who don't have a
technical background that just want to
be crap you know kept abreast of the
news you know I'm trying to do like you
know a more technical audience but even
within a more technical audience there's
so many different like niches like I
think one thing that would be really
cool that I haven't seen too much of
um maybe to some degree is like a media
Outlet like specifically for AI tools
just like for developers like like just
like and you keep it really simple but
just keeping keeping software developers
and machine learning Engineers like
abreast with just like you know the best
tools that they could be using in their
in their work like even something like
that would just be so extraordinarily
valuable
um so I'm not doing that but I think it
would be really cool if someone did yeah
it's crazy how
I don't know if it like it wasn't like
this I feel like in with the internet in
the beginning but it's crazy now how any
very Niche topic can blow up and work
really well sure no I think we're both
on Twitter like one of my favorite
accounts is this guy called strip mall
guy I think he's like kind of famous on
Twitter and he's just like he's this guy
and he's he only tweets about like
buying and flipping strip malls and like
renovating strip balls and he has like a
hundred thousand followers or something
like that
um so there's there's people like this
all over
um
that's so cool you can like any passion
that you have or just something that you
do you can share about it and some other
person will certainly be interested do
you uh do you have like a favorite like
weirder Niche Twitter account I don't
mean it if you know I don't mean to put
you on the spot but I'm curious uh you
you kind of are because I'm I'm a really
bad twitter user so I I actually wanted
to like this was part of my questions
that I actually skipped but I I will ask
them so since whatever
um
for myself personally but also anyone
else that is that is listening because I
just give a small background to my
question it's basically all my friends
and the people I know like because I I
assume it may be because we are from a
small City and here it's French not in
English or something but nobody is using
Twitter like back in back in the days it
was Facebook and now it's messenger
nobody is using Twitter or even have an
account how would you convince them that
Twitter is useful or like why is Twitter
useful and why should should they be on
there
yeah I would say I would say two things
on a high level and it just depends so
some people don't have these interests
but I think the two things that are most
useful that are kind of separate Are One
Professional and then uh two are just
keeping abreast of news
um
you know professional like just the
networking that you can do not like a
weird sleazy way but just like meeting
people who are in your Niche
um is is just wild and it's just so much
better than LinkedIn like LinkedIn is
the site where you're supposed to be
doing this but like I don't know like
people don't really like hang out and
have genuine conversations on LinkedIn
or like people like I don't know it's
it's not it's much more it's it's much
more like genuine on Twitter
um so I think if you're interested in
doing that like uh you know you should
try it out uh and they're and they're
just crazy stories of like people people
who have made crazy professional
progress by just going on Twitter
um you know so I think there's
professional and then the other is just
like I think it's the best place to get
news uh across a variety of niches you
know
um you can just find all of these random
people who are like domain experts again
it's like you know I follow like Jan Le
and like Sam Altman but like I also
follow like strip mall guy or like yeah
olaji and it's like this really cool
like it's kind of like you craft your
own newspaper that you get every morning
from like all of these different like
Niche experts on exactly what you want
to learn about
um so I and and also just like
empirically like the the quality of
information you get
um is frankly higher and faster than
what you'd get out of traditional media
um you know I think with like a lot of
political issues uh things like kovid
um you know thing things like inflation
at least in the US like Twitter at least
I found was way ahead of the curve
um on like finding up-to-date info about
this stuff
um
so I think just even keeping abreast of
what's going on like Twitter's great
now once you you spend the time to craft
your your list or like the people that
you are following this seems like a
really good tool to to stay in the farm
in as you said I'll Niche I'll
most important
subject for you so definitely and coming
back to just before we I forget
coming back to your AI Pub the the
Twitter account that's this specific
tutorial how how would you say that
like higher how are you balancing
the need to share super Cutting Edge
research like in your weekly reviews
well also need to communicate the need
to communicate them simply and make it
accessible like how is it for you is it
difficult to share the most advanced
algorithm while making it accessible and
all this and
Twitter it's like basically really short
threads so it what are the challenges
and how are you doing that yeah it's
definitely hard I mean the biggest
challenge is just volume
um and I don't just don't even pretend
to be a source of like all of The
Cutting Edge stuff it's more I feel like
I'll pick out a gem every once in a
while and maybe it helps people
understand it but it's like there's just
even if I was doing my job like 10x
better like I just being able to cover
just all of the things that are coming
out is just like really really hard
um so for me like the biggest challenge
is just volume
um and also Pace like just you know a
paper will come out you know
you know Day end like if you can
figure out what's going on in it and
summarize it on Day N plus one like
that's that's great but sometimes it's
just hard to do
um
I actually like in some ways I don't I
don't actually find there's a ton of
tension between
um
you know there's this question like kind
of explaining the research versus you
know crafting a Twitter thread like like
in some ways like you're kind of
compressing information and how much
information is lost when you do that
yeah um for me I don't I don't find
there's like a huge trade-off there
because at least what I'm expressly
trying to do is is kind of trying to do
some of that lossy compression uh you
know so for me like what I'm trying to
do is trying to kind of distill as much
as I can
in the paper into something that could
be a Twitter thread or could be
explained yeah you know a page or two of
writing um I think for other people who
are going much deeper into the technical
weeds or like are you know trying to
explain things for people who have much
more of a background I think there's
more of a trade-off but uh for me for me
not as much and that's kind of why I
like Twitter like I like that there's
at least to some degree in the UI like
the 240 character limit because it
really actually imposes a constraint of
being of being succinct and pithy yeah
and like just really summarizing like
the key ideas
um
so I don't actually find a huge tension
there personally
it's really funny actually because I
feel like oh sorry I I feel like I feel
like for people who don't use Twitter
they're historically has been this
reputation of Twitter as like this kind
of like
I feel like low IQ yeah like platform
it's like oh yeah it's just 140
characters like what kind of
intellectual content are you gonna get
there and it's like surprise actually no
it's like one of the best places to
learn
um so so it seems like there's a trade
off there but but at least I haven't
found it personally
um
yeah indeed and making stuff simple is
also a super useful skill for just
Professionals in general so would you
have any tips based on how you
personally do that to for anyone to
improve on explaining their research or
their own work more simply
well that's hard
um well the first thing that's kind of
obvious advice is get on Twitter
um it's really funny it's like a lot of
the research that ends up getting the
most attention or like most follow-up
work is just the research that gets
shared on Twitter because everyone's
using Twitter as their source of info so
I really the first and most obvious
source of advice is like just go on
Twitter and like every time you publish
research like dedicate a good deal of
time to making like a really good
Twitter thread with you know photos and
media and stuff uh that explains what
you did at a high level
um
was in terms of summarizing
I don't I don't feel like I have like
super interesting thoughts there it's
like you can write short sentences kind
of space it out like you know
yeah well I think those are already
great insights just basically I think
it's it's interesting to force yourself
to cut some words like make it more
concise by intention this will just
allow you to practice your brain to to
even verbalize things more simply in
their General Life just because you you
are doing that you are forcing yourself
one more thought here which is
um
trying to condense something to like
three major ideas or like trying to
understand why something is obvious it's
I when I was an undergrad I was like
really interested in doing math and I
was I was mentored by this like really
awesome mathematician
um and like one of the pieces of advice
he had like even for like extremely
technical work these are like math
papers which are like way harder to read
than than machine learning papers this
is totally esoteric like very awfully
put together often
um his advice was often like there are
three major ideas and like your goal
reading this like really hard esoteric
thing is to just figure out what those
like three major ideas are so like kind
of deliberately approaching your
explanations or your
uh reading from the perspective of like
what are what are the three bullet
points that explain this whole thing or
the other the other kind of like useful
intuition is like why is this obvious
like because the funny thing is like all
this intellectual work like some really
interesting paper comes out sometimes it
seems like this incredible technical
feat or like this leap of Genius of the
mind
um but but the person who discovered it
it was obvious right like to the person
who discovered it was actually very
intuitive so then sometimes I feel like
the goal of either coming to understand
the stuff well or explain it well is
trying to like figure out like trying to
reverse engineer the headspace to the
point where the work becomes obvious so
you're kind of asking yourself the
question like why is this really
intuitive even if it seems hard like
what's the intuition behind it um
these are two ideas I learned from math
that yeah they are good tips and also
what you mentioned with using external
media like images or videos or even just
real world example just like something
that anyone can understand but you can
compare that's extremely good to just
make things sound simpler and more
understandable I I think well most
professors do that they start to most
good professors do that they start with
a real world example or real life like
they talk about tennis or whatever and
then they talk about math
whether you like it or not a lot of this
stuff is marketing you know like you
know you you do the research that's
making the product but then in order to
get people to pay attention to it it's
marketing so you know learning the
basics of copywriting and all these
network marketing things that a lot of
people are reverse to it's useful
exactly and just just like selling
yourself in an interview or anything
it's it's all about marketing yep so you
you mentioned that you were
mainly staying up to date with Twitter
and
yeah Twitter
and so is there since you are yourself a
Twitter account trying to share news in
Ai and what my knee and AI
are there any other tools that you are
using to help stay up to date or
newsletters or I don't know YouTube
accounts or anything else than Twitter
accounts or if if there are other useful
Twitter accounts that you can share yes
um okay so you don't want to like play
favorites here but I'll and I I'm I'm a
little bit sleep deprived today so I'm
sure like some great accounts are gonna
skip my memory um nonetheless uh a
couple that come to mind
um Dar AI or I don't know how you
pronounce it but their AI is like
they're really really good
um they I feel like there are just
actually a lot of Twitter accounts that
are now just doing a better job at what
I'm trying to do than I am
um so I think I think like Dairy eye is
really really good uh they're doing
weekly paper summaries which is like
super cool so if you're trying to keep
up to date with research like I think
it's every Sunday they publish a thread
where it's like hey here like the top 10
like papers that came out this week and
here's like one or two sentences about
like what they contain
um so dare AI is really good
um Alpha signal you know uh Weekly
Newsletter with kind of like some of the
top AI papers that's really cool
um
uh I really like uh Dr Jim fan
um he's he's on Twitter he like he makes
like really cool explainer threads of
various AI research stuff
um and then and then another uh kind of
Follow that that I really like that's a
little bit different from the research
stuff
um is is this podcast called the
cognitive Revolution that I mentioned
um is run by uh this guy Nathan lebens
uh who's just a great guy and uh it it
covers some AI research stuff but also
just like how AI is affecting society
and they like interview some of the top
like startup Founders or like VCS in the
space and also researchers so it's like
more of a kind of like broad uh like
podcast uh that that I think is good and
they also they also have like a sub
stack or a newsletter not exactly sure
on that or they they do have one I I
mostly just listen to the podcast
I I will also add just it's basically
this the only with yours and I follow
Alpha signal AI Pub and one other
Twitter page account
um for this is mainly interesting for
people enter in into research AI
research and
even more towards
like researchers applying those research
but it's obviously AK oh yeah yes you
already know but this this was a game
changer to me like a few I think years
ago now well and it's definitely cool he
posts he basically shares
every new relevant papers and even
non-relevant but like any new papers
with a short one-liner that is usually
taken from the abstract but still a
short useful one-liner to know what's
the paper about with a picture and it's
it's just really cool to to be easily to
have easy updates on on the new papers
the new like thousands of daily papers
so that's that's a really good one I
feel like yep I totally agree totally
forgot uh another person who's in a
similar vein is Iran Iran kamat suzaki
um he's he's at he's doing a machine
learning PhD at Georgia Tech I think
he's finishing up uh and he he does
similar stuff like he posts a lot of
papers and kind of summaries and papers
uh awesome
oh I'm I'm sure you do uh he's he's very
active on on AI yeah I must have seen it
but I my mirror is extremely bad anyway
so I probably know know him but yeah to
continue on on Twitter since uh this is
the pretty much the main topic of the
past 30 minutes or so I I have still
follow-up questions on this and it's
also for people like me that maybe do
not understand Twitter enough
and it's more about growth so
a very short question that may be
extremely hard to answer is how to do a
viral Twitter post
oh that's um that's what I'll put I I
don't mean to toot my own horn too much
uh but I do actually think I've Gotten
Good at this
um especially yeah yeah now I've been
actually kind of like slammed with some
of this recruiting stuff and other
things I'm doing so I actually haven't
been focused on growing my Twitter
account but yeah like first first couple
months it was like very deliberate very
focused on it and I think it did a very
good job
um
yeah I think so there are a lot of
thoughts
um how do you write yeah so there I mean
and maybe this is too vague or too broad
but I think you gotta get like three
things right I think one getting the
audience right like kind of it's kind of
like finding product Market fit it's
like finding like info audience fit like
like like being really deliberate like
think like a think like a business like
ask who are your customers and ask like
what do they actually want like what
would help them like what do they
actually care about rather than like
what's exactly interesting to you is
because often like those very subtle
differences in that
um so like kind of like finding like a
niche audience or once you have an
audience like figuring out exactly who
that is and like really being deliberate
like what did what would actually be
very useful to them
um you know and um you know thinking
about that
um you know
um I think so so that the second is is
content like finding content that's
interesting and relevant
um I think one thing that I definitely
found is posting posting like long form
content
um that's like really thought out and
valuable like occasionally does much
better than
um really frequently posting content
that's not necessarily like stuff that
you have added or like reposting stuff
for just like sharing a random things I
think like there's a huge return like
one of the reasons I actually don't post
too much on Twitter is because like
often it takes me like several hours to
write a single thread because it's like
I'll read the paper like come up with
some of the major ideas and all that
like it takes a long time
um so I think like you know figuring out
what your audience wants then like doing
really like long-form valuable stuff in
that domain uh that's like detailed and
helpful
um and then the third thing you kind of
have to get all these right too uh
because I feel like if you get one of
them wrong like you you it really you
don't grow as much or you don't get get
as much um
views
um the last is just like kind of
straightforward like marketing
copywriting stuff
um so it's like you know just the way
you structure your sentences or like the
way you structure your tweets like each
you know have each tweet like not have a
ton of text in it like have each tweet
ideally just contain one idea
um you know structure your sentences so
that they're like short easy to follow
structure your sentences so that they're
kind of interesting so it's like you
have one idea and then the next idea is
like a plot twist or like you know uh
you just you don't want it to be too
monotonic right
um you know and then also including
media like including like images for
each each uh each tweet like these these
kind of subtle things actually end up
mattering a lot so I think
maybe that's not super explanatory but
that's kind of my like high level view
it's like you get infinite you know
audience info fit think about what your
audience actually wants like think about
them like a customer how can you serve
them uh you know long form detailed
stuff and then like doing the
copywriting and the and the images and
stuff
and how much time did it take when you
were trying to really grow on Twitter
like per day or per week
approximately it really depends like I
wouldn't say actually a ton of time
honestly
um I think like I think when I was
starting maybe like 10 hours a week or
something like that
um because I was doing well actually
maybe more than that because I think
like when I was starting out even when I
was starting out and I was growing
really fast like I wasn't actually
posting that much stuff I was posting
like twice a week basically like I would
post
roughly every week like one long form
explainer thread on like some cool paper
or some cool machine learning topic
um and then I would do like a weekly
like uh summary of the stuff that
happened in AI or AI research uh and
each of those okay maybe it's actually a
little bit over 10 like the the summary
thing I think even to put the thread
together would take like a couple hours
uh but then throughout the course of the
week I would just save all of the links
that I found on Twitter to like filter
through that so maybe it's like 15 hours
something like that and then and then
you know uh writing a thread on a paper
like
maybe it's like seven hours or something
like that like you find the paper you
print it yeah I'm boring I you actually
read it on paper and Mark it up
um you know that can take a couple hours
to really understand and then um you
know writing the thread and doing all
the like the copywriting like I think a
lot of people just don't like doing this
because it's actually kind of boring
like just getting getting the images
right and getting the sentences right
like that actually takes a good deal of
time yeah I think for me it was like 10
to 15 hours a week something like that
and right now how long does it take you
I think I'm actually spending maybe a
similar amount of time because I'm also
doing a podcast now
um
maybe actually maybe a little bit less
than ten something like 10 I would say
something like that
yeah that is uh it's not really that
many hours to get that much views and
and just so many opportunities oh yeah
yeah less than we think total life hack
like like even you know someone who's
just you have a full-time job you know
you do a couple hours a week like and it
just it totally changed your life
indeed and you mentioned cool research
papers but how do you decide if they are
cool or worth sharing you you said that
you have to see as if you you were
basically sharing with customers but
it's still
you in the end you still have to decide
if you are explaining a paper or not so
how how do you take these decisions yeah
I think in terms of finding the papers
and maybe you're just a little bit more
selfish I think I'm actually thinking
less from the audience perspective and I
but it's it's more like I'm just using
my own taste or my own judgment it's
kind of like with even with math like
people just have different tastes you
know some people like analysis other
people like number Theory or something
you know
um and with machine learning papers like
it's just kind of like what I what I
find most interesting and typically what
I find most interesting is like
I would say less like really cool demos
or something or like a really impressive
state of the art result or something but
I think what I find very interesting is
like papers that open up new paradigms
or papers that
um kind of have something to say about
the long-term direction of uh AI
research I also like to some degree I'm
in kind of an AGI believer like maybe
even short timelines type person so like
I also find papers that kind of like
play into that thing interesting you
know so it's like
uh some of the papers that I've covered
recently on the podcast that that I
found really interesting or like tool
former so it's like okay that's like
just directly Society societally
relevant um or like you know recently
did one on um Hungry Hungry Hippos which
is like a new language modeling
architecture that like has much better
scaling with um context uh the context
length so like stuff like that like
where it's like oh if you get that right
then the language models are gonna get
way better
um so it's mostly it's mostly my taste I
think I'm mostly interested in language
models
um or like or or like Foundation models
that incorporate language so I think
there's also a lot more bias towards
that than like vision
yeah you you definitely have a bias Just
Like Me on YouTube I I share what I find
interesting and what I like it's and but
I think that's a good thing it's also it
also gives personality to your Twitter
account and if people enjoy what you
seem to be sharing well they won't be
disappointed since you will keep sharing
similar stuff
yeah yeah there's always I think this is
kind of kind of interesting Trend in
like media there's always this like
interesting tension that I've seen
between
um
kind of doing like writing for yourself
or producing media for yourself that's
like unique to your character and
like sort of serving your audience as if
you were a business or a servant or you
know kind of like a more of a giving
mentality
um because you kind of have to do both
like if you're just so obsessed with
like what you think is cool like
actually often people don't care like
it's just not interesting to other
people but then on the other side like
there's definitely a phenomenon among
like influencers of like audience
capture where it's like they start doing
something that's really interesting and
unique but then they find that Their
audience actually wants more clickbait
and then like over time they end up just
becoming like a clickbait Channel or
they end up just being completely
transformed and captured by Their
audience so it's
it's definitely something that you have
to like think about yeah and speaking of
audience how how are you engaging with
your Twitter audience or just building
some kind of real engaged Community
rather than just being a anonymized
Twitter account that doesn't talk to
them do you do anything more specific
than sharing your your weekly posts or
all your posts well I'm doing much a
much worse job than you are because you
actually run a community I I'm I I feel
like I could actually improve a lot in
this respect I think insofar as I'm
doing anything community
oriented I think it's actually much less
like scalable
um or it's very like I think I'm
actually doing more like in-person stuff
so like yeah occasionally all host
events you know I feel bad because it's
like they're NSF like if you're not NSF
but like you know a whole host events
are all sponsor events that are like
open to anyone who wants to show up
um and then some of the some of the
recruiting stuff uh is also like useful
so like a lot of the stuff that I've
been doing community wise is like in
person where it's like there are just
people who I've been chatting with in
the DMS for like a couple months and
they're like looking for a job in this
area and like I'll help them out
um or just like connecting people but
it's very it's very like non-scalable
like relative to I don't know
do you have any plan in in creating
something that is more oriented towards
some kind of community and where people
can
help each other or you can help them
more easily than in person because of
course in person is like the best thing
but it as you said it's definitely not
scalable
to be honest not really like right right
now uh I feel like I'm still like
figuring out all this recruiting stuff
like a little bit slammed uh just with
all the things that are going on so I
don't
I think one thing that I've tried to do
with the account is like if I'm doing
anything I want to do like really good
jobs so I I don't feel like I do a very
good job with the community thing right
now
um so there's nothing on my docket
yeah that makes sense you you are just
prioritizing and focusing on other stuff
that is more important to you and that's
that's just perfectly fine and well
speaking of managing the the Twitter
account is there any challenges or
difficult things that we people that are
not really good with Twitter or are not
managing such a popular Twitter account
uh yeah sorry I will just before it are
there any
challenges that we do not expect to see
from managing oh
something that maybe you didn't expect
at the beginning yourself
that is that takes you more time than
expected or yeah I mean it I think well
I would say two things
um one is it depends on what kind of
account you're running like if you're
running a newsletter for example I know
people are running like daily AI
newsletters you just I used to you know
yeah you have to keep up you have to
every day put that out like there are
people who are depending on you yeah I
feel fortunate enough that like kind of
what I'm doing can be at my own pace so
it's like if I'm really slammed like
you know I'll just you know I will post
something about a research paper that
week yeah um so I I don't have as much
of that pressure but I think
um you know other people would
experience that
um
I I think the the main thing for me
that's also true of like some of the
recruiting stuff is just like inbound
and just like DMs I feel so bad I'm like
I and it's it's also a skill so it's
like it becomes hard but then you get
better at it
um I still like not good at it but it's
just like you get a lot of like inbound
and a lot of people are like oh can you
share this and I'm kind of like yeah I
don't really know like just lots of
people who want to share things to you
or like do you want to sponsor this
event or like with recruiting too the
funny thing is like recruiting is like
it's a pretty it's pretty unsexy
business but there aren't actually like
that many people it's so funny it's like
there's so many AI startups so many AI
VCS really not that many ad recruiters
like I kind of don't know of any
um so then if you start to get a
reputation as like the guy who does AI
recruiting like a lot of people will
just come to you and they're like can
you send us candidates or you know uh
so just kind of keeping up with that uh
I'm trying I'm trying to do a better job
that's really cool so you are a Pioneer
in this in this subfield a little bit
yeah it's just some degree
and that's also the main way you manage
to monetize your Twitter account right
were there any other ways that you you
succeeded in in making money through
that like that's such a big amount of
views that that you get
um that's the only way I've done it so
far I for a while I started I started a
job board I had like a job board yeah
what I found is that I don't know like
because again I kind of was talking
about like how recruiting is pretty
non-scalable and like I would recruit I
will recruit for startup CEOs and
they'll tell me that they just don't
even look at their job submissions
anymore so then I was kind of like man
like even if you could make money doing
this like are you really helping
um you know so I had this job board for
a little bit but I kind of just wasn't
clear if I was actually getting anyone a
job from it
um so that was the main way of kind of
approach monetization
um I don't know if it's the best way
um recruiting especially in AI is is
very high ticket you know if you charge
a 20 placement fee
um and someone gets paid a salary of a
hundred and fifty thousand dollars is
thirty thousand dollars uh so it's it's
the the ticket size is very high
um the the one thing is is very
non-scalable so it's like I just spend a
lot of time like sending people emails
or like pinging Founders or like hopping
on calls with people whereas I think
like a lot of the ways that people tend
to monetize media
uh via sponsorships or like ads or
things like that are like much more
scalable right it's like if you have a
newsletter or someone just like clicks
something to post on your job board or
like post it yeah you know something
it's just it's it doesn't take an
additional time
um on your end I think if your audience
is interested or like I have thought
about this stuff like I do think about
this like a business and do I do
actually think about making money
um one thing that I think you just
make tons of money from that I haven't
experimented from but I've seen other
people do is like events um there's just
so much demand for events in the space
like hackathons or things like that uh
real world events right because I've
I've seen like a drastic drop in online
events attendance just well just for
myself I'm not really that interested in
online events anymore I think we we've
just all did a
uh I forgot the word but we we had way
too much online events during covid and
everything so yeah I I think a lot of
people are super into real world events
now even even more than before Colvin
maybe and I I assume that's what you
meant by events and hackathon like
mainly in person yeah I would I would
say in person and like I'm not an expert
here and like I'm so busy that I'm like
not not doing this personally but it's
like the economics are really
interesting you know because it's like
you you can like even a hackathon you
don't charge anyone any thing to show up
like often you can get the event space
for free from a sponsor and then you
have like 50 people who will do 50
different companies that sponsor the
event for like two thousand dollars or
something it's like surprise you've made
a hundred thousand dollars in a day you
know
um then I so I've just seen this like
events space and I've started to talk to
some people who like do this
professionally and it's like really
interesting yeah also I went to I don't
know if you know about ai4
oh yeah yeah yeah yeah yeah it's an
in-person event with lots of
professionals and these events the
tickets are like extremely expensive
it's for companies to basically learn
about other companies and connect and
like find Opportunities and it's I'm
pretty sure it is it it makes a lot of
money I I don't know personally but it
seems it seems quite quite beneficial
yeah
and okay so my last questions about
Twitter then I will just conclude with
some other questions about your other
projects and and you but
um so Twitter we all know about like
Elon Musk and Trump and everything going
on with Twitter so yeah do you have to
handle some controversies or like
polarizing topics or and if you do how
how do you handle those or can you just
completely like not respond and not not
do anything about that
I think at least for me the the niche
that I'm in I don't have to engage with
any of that stuff which I'm I feel very
fortunate about
um and I actually try to avoid it
because it's like I you know I think a
lot of people are following me for like
AI research so it's like
if I started to post about political
stuff or start liking political things
like I just think people wouldn't like
it so I'm I'm very fortunate I think
like people are following me they want
AI research they don't you know they
follow other people for you know Trump
or Elon
awesome and that definitely makes sense
as well I I don't have any issues either
but I'm also much smarter on Twitter I'm
just basically sharing my videos so it's
not really a good way of using Twitter
but anyways I completely understand that
and my
I wanted to talk a bit about your
podcast but you already mentioned what
you are doing on there and I I have I've
also listened to the recent one with uh
the two the two people I don't remember
the name but the two people from openai
which was really cool yeah yeah
and uh yeah I was just okay I wanted to
talk just a little bit about the podcast
because I'm now hosting one as well and
just maybe that that could be
interesting to some people because
um we know that podcasts are becoming
incredibly popular like a lot of even
myself everyone is listening to podcasts
I assume you are too well you mentioned
that you were listening to One so
we all are listening to to podcasts and
so some of our audience may want to even
create one themselves if they are
interested into that that's also a good
way to to be online to have an online
presence
and so
what's to you the most challenging thing
about hosting a podcast
oh I don't have a good answer here
because I feel like
um I have a very fortunate setup so the
the podcast was actually proposed to me
by uh the CEO of this model modern
monitoring company a rising a guy so
this is this guy he's he's crazy he runs
this huge AI company huge AI startup
he's you know working around the clock
he somehow he also his name is Jason
lopateki he also manages to find time to
like just read AI papers like like I
don't I don't know how he does it but he
reads AI research like on a day-to-day
basis so he he actually reached out to
me and he's like hey let's do this
podcast
um so for me it's actually been very
easy
um because he has a marketing team where
basically we do the recording but then
um you know his marketing team is
absolutely excellent and they'll handle
something like the editing and things
like that I think for me the hard the
hard part or the the typically the hard
part with doing podcast and media is is
just a lot of the editing stuff and the
the marketing things like that
fortunately I do not have to handle that
um foreign
on my end like and the funny thing is
even the the format of the podcast is
not that hard because it's I don't even
have to talk too much because I'm just
I'm interviewing you know AI researchers
so just ask them about the paper so the
really the only hard part for me is just
is is reading is actually reading the
papers uh but that's that's pretty fun
for me uh so so typically I'll read you
know read the paper in advance and come
up with you know some high-level
questions
um I've actually I was like really
hesitant to take on the podcast because
I I didn't I wanted to do a really good
job with it and I thought it would be
really hard but it hasn't been super
hard I think I think I think I want to
increase the volume again you know we're
doing something like once every three
weeks or something I think a lot of
really good podcasts are doing like once
a week so I think if we could up the
volume that would be cool
yeah that's what we call I'd love that
as well myself and
what you mentioned that you you like
doing that and there are no like
there aren't big challenges for you just
reading the papers that takes time
obviously and and coming coming up with
the questions which I for a tip for
anyone that hosts an interview I think
that like of course you need to draft
some questions and think about the the
person and and do some research but also
I've done that in the past it doesn't it
is the vast major minority of my
questions but I have also asked
some AI is like charge GPT to generate
more questions given the background of
someone and some question examples it's
a really good idea yeah it's for for
real it works pretty good like it's it's
quite repetitive but if you ask if you
ask it you will definitely go it will
definitely come up with at least one
interesting question that you haven't
thought of so definitely was gpt4 in the
long context length like yeah I see I
feel terrible I actually I I haven't
used these tools like
a ton in my day-to-day work but it's
just the value-add is so obvious like
yeah especially with a long context
length of gpt4 like you just copy paste
like a whole paper in it or even just
take a screenshot of the paper and just
say like hey what are some cool high
level questions that you have about this
yeah yeah I just started using it well I
I'm using it more and more now it's it's
funny because it's my PhD Professor that
is pushing his students well us to to
use GPT more so it's it's really it's
crazy I I wouldn't think that a
professor would actually recommend using
an AI to to write and code but he's
doing that and it's it's extremely
powerful of course you need to double
check and like be sure that it's fine
but it's really helpful especially for
me that and other people that are not
native English speaker it's so good to
just like for example I write an article
and just and then I just copy paste it
and ask to keep the information the same
but rephrase it and it transforms it
into like a very professional article
written by like a perfect thing with a
perfect English this perfect grammar
everything so it's that's pretty cool
this is like truly truly magical
technology that's that's so wild yeah
and
yeah what what do you enjoy most about
hosting a podcast
I think it's either I just like talking
to people I think I think it's really
cool to like talk to the people who are
actually figuring this stuff out or
actually building these systems
um you know I just I feel very fortunate
that you know I can kind of uh at least
make a living uh
understanding and following a lot of
high-level trends that really matter to
me so just like just feel very fortunate
to like
be able to at least spend a good chunk
of my time like reading papers and
thinking about it
um
yeah I think I think the coolest thing
is just being able to follow these
Trends um and also like talk talk to the
people because they have like really
really insightful things to say yeah and
they're often really good communicators
I I feel like especially like the last
episode
um the last episode on tool former just
put it out like this week like
especially Teemo like Teemo just like
gave a really really good explanation
like he was just really good and kind of
lively Communicator so I think this is
also just fun are you planning on
interviewing other kind of background
and researchers with their research
papers
yes I don't actually know who I'm gonna
do next I don't know who's next I I want
to maybe do something that's not just
language models because I I I'm I'm
really into this stuff but I feel like
maybe I cover it too much
um
I it's like obviously the the next
obvious candidate would be gpt4 but if
you read the paper they don't really
want to talk about what they did yeah
yeah so I don't I don't know if they're
the right people to to interview
but all right I also want to say I don't
mean that I I don't uh I don't want I
don't mean to say that to like knock
open Ai No no of course of course it's
incredible but no it's it's really
incredible what they did and yeah
obviously and I just I mainly wanted to
ask if you were planning to to change a
bit your interview is like not only
cover research papers but maybe maybe
get people from startups or
anything different in the AI world or
will this just stick you will stick with
research papers for this particular but
yes
for this particular podcast probably
just stick with research papers again
there's just this whole thing where I I
think the kind of the AI media space is
pretty underserved
um there's just there's so much that I
would like to do that like I don't
exactly have bandwidth to do like I
totally I think there's so much more to
do in the space I don't mean to like
show this one podcast too much but I
just do really like them I do think the
cognitive Revolution is doing a really
good job there
um but yeah I like I think there would
be a ton of cool stuff just to like
interview people about like machine
learning careers like just like
interview people who's like uh machine
learning Tech lead or like a machine
learning manager and talk about like how
did they get there like what's what
actually matters like in terms of
learning on the job or getting a good
job
um there's there's there's one guy who's
on uh machine learning Twitter I I hope
I get his name right or like I I'm not
sure if I exactly remember his name but
I think his name is Alexa gordick uh
he's he's a deep mind I think he
recently is taking a break from deepmind
but like his story is really cool where
he kind of took a an alternate path to
get into being a research engineer at D
mind and like doing an interview with
him would be like super cool just like
how did he do that how do you pull that
off and they did yeah
you're I I will actually have it have
him on the on the podcast and he he's
taking a break in Japan I believe for a
few months and then in I believe in may
we will have a podcast together so oh
cool yeah he has a great background
um okay speaking of uh GPT generated
questions I think this is the question
that GPT gave me and I I really liked it
I had to rephrase it but it it had a
really good ID and so I will read it
just to keep it keep the credit to GPT
um since you are really active on there
how has social media impacted the way AI
research is communicated and shared with
the broader public and I can like maybe
spend
um
specialized to Twitter like how how
would you see Twitter how are you seeing
um
modifying the way resources shared
compared to the Past
yeah I mean I think
that's an interesting question I think
you can comment Less on how the research
gets shared to the public because I just
kind of don't know
um
I don't know if I have anything to say
that's not kind of obvious I mean one
one thing that is just totally true is
there's kind of this formal reality of
how AI research gets disseminated and
then there's the actual reality like
there's there's all these conferences
you know
um all these journals and stuff but then
you talk to any of these researchers and
they're kind of like none of this stuff
matters anymore it's like it does matter
especially if you're trying to further
your academic career or you put you know
or iclr on your resume or something but
um insofar as like what are people
actually using to keep up to date like
the top researchers in the space it's
just Twitter
um it's really like it's the place that
people are using to uh
to learn about what's going on
um
so I feel like that's kind of my only
observation is like it's actually even
even if it's not formally the mechanism
that people are using to aggregate a lot
of the most interesting work it's
I think factually what people use
um and then as a result like I do think
you just get
you can totally change the impact of
your work by just focusing on marketing
on on Twitter and there's there's
there's kind of a hard effort there's a
weird ethical line to
tread here where it's like maybe you
will increase the breadth of your kind
of the the extent of the replication of
your research by you know making lots of
like click-baity threads about your
research or like making lots of threads
that kind of overstate the the promises
the results of your research I think
that's true
um but nonetheless I think it's like a
lot of the research that really gets um
gets followed up on is is the stuff that
does what does does well on Twitter so
if you're an academic I I think it
actually is quite imperative if you want
to you know yeah if you actually think
your work is worth following up on to to
like you know make Twitter threads about
it I think
yeah there I I don't have the reference
but there has been a study as well where
like they shared the researchers shared
on Twitter the half of the papers and
didn't share the other half and it
clearly had an impact the Twitter paper
clearly had an impact in the reads
obviously but it even had an impact in
the number of citations the paper had so
that's
I don't know if it's a good thing but
it's definitely worthwhile to share it
on Twitter if you have published a paper
but yeah speaking of a good thing or not
would you say that Twitter has a
positive impact on the research and
their research industry like the
that's really hard to say I don't think
I'm an expert like I did a year of a
machine learning PhD I definitely can
read and understand these things at a
high level but like I'm not you know I'm
not you know a 10-year Professor I'm not
no someone who's done really world-class
research at deepmind or something
um so I don't actually feel like I'm
qualified to comment
um I think the one like meta comment I
would have is generally in life I don't
think of things as being like single
valued but I think of them as being kind
of like multi-dimensional or vector
valued so insofar as I wouldn't know if
it's like good or bad but I think on
various Dimensions it would be good or
bad so it's like one dimension would be
like yeah probably the more clickbaity
stuff like gets shared more and people
believe in it like maybe that's bad
um but then also like um it's very uh
it's in many ways very democratized like
like truly if you're like on the other
side of globe and you do cool research
and like you write a cool Twitter Thread
about it like people will see it um and
like that's that's really cool so like
that would be one area where I think
it's really good
it definitely lowers the barrier to
entry like a lot of these kind of old
boys clubs uh you know
uh don't matter as much anymore of
course sir for sure
and so my my last question
is you you already talked a lot about
this so I assume we already know the
answer but still I I want to
to ask it it's
basically well it's it's twofold very
simple it's just what is your next
project or like current Obsession that
you are working on and also what would
you like the people to know or follow
from you
um next project what I'm working on I I
feel kind of weird it's not super
related to Twitter but it's kind of an
outgrowth of the recruiting stuff is um
there's this one company I'm really
excited about they're doing really well
they're trying to hire like top one
percent Talent um trying to be careful
about what I say about them so I I won't
mention who they are um but
you know if things work out I think uh
I'm gonna not like work for them as an
employee but like uh partner more
closely and just like try to fill a ton
of their roles um so that for me is like
a kind of interesting experience where
it's uh it's more high pressure and it's
just definitely like
it's different from what I've been doing
in the past because in the past I've
just been this like passive Matchmaker
like people show up and I'll be like oh
I think you'd enjoy this job whereas now
it's like okay here's the org chart like
here's all the roles we want to fill in
the next like
you know four months or something like
let's do it you know um so I think I
think that's one of the biggest projects
that I'm approaching right now that I
think will be really fun and and really
hard and really new to me
um
and what what to follow me on Twitter or
uh or what it is yeah would you like to
mainly share your Twitter account or
maybe well this is undisclosed so you
cannot really talk about it that much
but that much but like I assume AI Pub
is the best place best place to to learn
more about you and just follow your work
but is there would you like to share
your recruiting platform or oh sure just
oh yeah sure yeah I mean you it's all
kind of via Twitter like I don't have a
newsletter I don't have a website or
anything so if you just go to my Twitter
page you'll find everything uh you know
that is it's all very simple like you
just click on the link and it'll take
you to the little form just to submit uh
you know your preferences for for talent
referrals
um I have a personal Twitter account but
I don't really actually put much effort
into it like I'm actually really
deliberate like on my personal account I
just feel like I can post whatever and I
don't have a high bar and a lot of it's
just like totally random personal stuff
it's like experimenting with sleep or
Fitness or things like that or oh I
don't you know I don't do this as much
anymore but I'll like post about books
I'm reading uh I was reading some like
interesting math textbooks in the summer
so it's like posting about that I'm like
reading an interesting book on
negotiation right now so is it the the
one from the CIA agent or no I will
maybe I maybe this is the same guy it's
it's actually a really good book it's
called Uh the the title sounds very
Machiavellian the title is um bargaining
for Advantage and then oh I think yeah
yeah this the subtitle is um negotiation
strategies for reasonable people the
book is much closer to this subtitle
where it's actually it's very like civil
yeah about being a good person and
maintaining relationships and like that
but it's really good I I'm enjoying it
okay then I I
forgot the name obviously but there's
one called negotiate something and it's
by an xcia agent and it's like I think
it's the best book on negotiation and
it's just applied to the real world
getting a job your your personal life
anything it's
incredibly good it's a really it's a
really good book so it's I'm just
opening a tab I'm just curious this is
yeah
yeah negotiate and CIA agent you you
should you should find it cool
I think just pulled it up never split
the difference yeah exactly that's
that's the one yeah a really good one I
certainly recommend but awesome so I I
might follow your your private account
just because I'm really into better
sleeping habits and and Sports in
general and just eating and everything
like for well
uberman and everything related to that
I'm I'm super into into this and I'm
also reading lots of book and you seem
to be
reading similar format as as I enjoy so
well anyways I'll definitely check that
out as well and I I really invite anyone
interested in AI a little bit I assume
you are if you are still listening to
follow AI Pub on Twitter it's a really
amazing account and I personally love it
and thank you very much Brian for your
time it was amazing to learn from you
and chat with you and yeah a huge thanks
for for taking the time to to be there
no thank you this is like super fun
really really enjoyed this so
[Music]
thank you
[Music]
( Spotify या Apple Podcasts पर भी उपलब्ध)।
पीएचडी करने से पहले इसे देखें। मशीन लर्निंग में | HackerNoon