back to indexKevin Systrom: Instagram | Lex Fridman Podcast #243
link |
The following is a conversation with Kevin Systrom,
link |
cofounder and long time CEO of Instagram,
link |
including for six years after Facebook's acquisition
link |
This is the Lex Friedman podcast.
link |
To support it, please check out our sponsors
link |
in the description.
link |
And now, here's my conversation with Kevin Systrom.
link |
At the risk of asking the Rolling Stones
link |
to play Satisfaction, let me ask you about
link |
the origin story of Instagram.
link |
Sure. So maybe some context.
link |
You, like we were talking about offline,
link |
grew up in Massachusetts, learned computer programming there,
link |
liked to play Doom II, worked at a vinyl record store.
link |
Then you went to Stanford, turned down Mr. Mark Zuckerberg
link |
and Facebook, went to Florence to study photography.
link |
Those are just some random, beautiful,
link |
impossibly brief glimpses into a life.
link |
So let me ask again, can you take me through
link |
the origin story of Instagram, given that context?
link |
You basically set it up.
link |
All right, so we have a fair amount of time,
link |
so I'll go into some detail.
link |
But basically what I'll say is,
link |
Instagram started out of a company actually called Bourbon,
link |
and it was spelled B U R B N.
link |
And a couple of things were happening at the time.
link |
So if we zoom back to 2010, not a lot of people remember
link |
what was happening in the dot com world then,
link |
but check in apps were all the rage.
link |
What's a check in app?
link |
Gowalla, Foursquare, Hotpotato.
link |
So I'm at a place, I'm gonna tell the world
link |
that I'm at this place.
link |
What's the idea behind this kind of app, by the way?
link |
You know what, I'm gonna answer that,
link |
but through what Instagram became
link |
and why I believe Instagram replaced them.
link |
So the whole idea was to share with the world
link |
what you were doing, specifically with your friends, right?
link |
But there were all the rage,
link |
and Foursquare was getting all the press.
link |
And I remember sitting around saying,
link |
hey, I wanna build something,
link |
but I don't know what I wanna build.
link |
What if I built a better version of Foursquare?
link |
And I asked myself, well, why don't I like Foursquare
link |
or how could it be improved?
link |
And basically I sat down and I said,
link |
I think that if you have a few extra features,
link |
it might be enough.
link |
One of which happened to be posting a photo
link |
of where you were.
link |
There were some others.
link |
It turns out that wasn't enough.
link |
My co founder joined, we were going to attack Foursquare
link |
and the likes and try to build something interesting.
link |
And no one used it.
link |
No one cared because it wasn't enough.
link |
It wasn't different enough, right?
link |
So one day we were sitting down and we asked ourselves,
link |
okay, it's come to Jesus moment.
link |
Are we gonna do this startup?
link |
And if we're going to,
link |
we can't do what we're currently doing.
link |
We have to switch it up.
link |
So what do people love the most?
link |
So we sat down and we wrote out three things
link |
that we thought people uniquely loved about our product
link |
that weren't in other products.
link |
Photos happened to be the top one.
link |
So sharing a photo of what you were doing,
link |
where you were at the moment
link |
was not something products let you do really.
link |
Facebook was like, post an album of your vacation
link |
from two weeks ago, right?
link |
Twitter allowed you to post a photo,
link |
but their feed was primarily text
link |
and they didn't show the photo in line
link |
or at least I don't think they did at the time.
link |
So even though it seems totally stupid
link |
and obvious to us now, at the moment then,
link |
posting a photo of what you were doing at the moment
link |
was like not a thing.
link |
So we decided to go after that
link |
because we noticed that people who used our service,
link |
the one thing they happened to like the most
link |
was posting a photo.
link |
So that was the beginning of Instagram.
link |
And yes, like we went through and we added filters
link |
and there's a bunch of stories around that.
link |
But the origin of this was that
link |
we were trying to be a check in app,
link |
realized that no one wanted another check in app.
link |
It became a photo sharing app,
link |
but one that was much more about what you're doing
link |
and where you are.
link |
And that's why when I say,
link |
I think we've replaced check in apps,
link |
it became a check in via a photo
link |
rather than saying your location
link |
and then optionally adding a photo.
link |
When you were thinking about what people like,
link |
from where did you get a sense
link |
that this is what people like?
link |
You said, we sat down, we wrote some stuff down on paper.
link |
Where is that intuition that seems fundamental
link |
to the success of an app like Instagram?
link |
Where does that idea,
link |
where does that list of three things come from exactly?
link |
Only after having studied machine learning now
link |
for a couple of years, I like, I have a...
link |
You have understood yourself?
link |
I've started to make connections,
link |
like we can go into this later,
link |
but obviously the connections between machine learning
link |
and the human brain, I think are stretched sometimes.
link |
At the same time, being able to back prop
link |
and being able to look at the world, try something,
link |
figure out how you're wrong, how wrong you are,
link |
and then nudge your company in the right direction
link |
based on how wrong you are,
link |
is a fascinating concept.
link |
And I don't, we didn't know we were doing it at the time,
link |
but that's basically what we were doing, right?
link |
We put it out to, call it a hundred people,
link |
and you would look at their data.
link |
You would say, what are they sharing?
link |
Like what resonates, what doesn't resonate?
link |
We think they're gonna resonate with X,
link |
but it turns out they resonate with Y.
link |
Okay, shift the company towards Y.
link |
And it turns out if you do that enough quickly enough,
link |
you can get to a solution that has product market fit.
link |
Most companies fail because they sit there
link |
and they don't, either they're learning rates too slow,
link |
they sit there and they're just,
link |
they're adamant that they're right,
link |
even though the data is telling them they're not right,
link |
or they're learning rates too high
link |
and they wildly chase different ideas
link |
and they never actually settle on one
link |
where they don't groove, right?
link |
And I think when we sat down
link |
and we wrote out those three ideas,
link |
what we were saying is, what are the three possible,
link |
whether they're local or global maxima in our world, right?
link |
That users are telling us they like
link |
because they're using the product that way.
link |
It was clear people liked the photos
link |
because that was the thing they were doing.
link |
And we just said, okay, like,
link |
what if we just cut out most of the other stuff
link |
and focus on that thing?
link |
And then it happened to be a multi billion dollar business
link |
and it's that easy by the way.
link |
Well, nobody ever writes about neural networks
link |
that miserably failed.
link |
So this particular neural network succeeded.
link |
Oh, they sell all the time, right?
link |
But nobody writes about it.
link |
The default state is failing.
link |
When you said the way people are using the app,
link |
is that the loss function for this neural network
link |
or is it also self report?
link |
Like, do you ever ask people what they like
link |
or do you have to track exactly what they're doing,
link |
not what they're saying?
link |
I once made a Thanksgiving dinner, okay?
link |
And it was for relatives and I like to cook a lot.
link |
And I worked really hard on picking the specific dishes
link |
and I was really proud because I had planned it out
link |
using a Gantt chart and like it was ready on time
link |
and everything was hot.
link |
Like, I don't know if you're a big Thanksgiving guy,
link |
but like the worst thing about Thanksgiving
link |
is when the turkey is cold and some things are hot
link |
and some things, anyway.
link |
You had a Gantt chart.
link |
Did you actually have a chart?
link |
OmniPlan, fairly expensive, like Gantt chart thing
link |
that I think maybe 10 people have purchased in the world,
link |
but I'm one of them and I use it for recipe planning
link |
only around big holidays.
link |
That's brilliant, by the way.
link |
Do people do this kind of...
link |
It's not overengineering, it's just engineering.
link |
Thanksgiving is a complicated set of events
link |
with some uncertainty with a lot of things going on.
link |
You should be able, you should be planning it in this way.
link |
There should be a chart.
link |
It's not overengineering.
link |
I mean, so what's funny is, brief aside.
link |
Yes, it's brilliant.
link |
I love cooking, I love food, I love coffee,
link |
and I've spent some time with some chefs
link |
who like know their stuff.
link |
And they always just take out a piece of paper
link |
and just work backwards in rough order.
link |
Like it's never perfect, but rough order.
link |
It's just like, oh, that makes sense.
link |
Why not just work backwards from the end goal, right?
link |
And put in some buffer time.
link |
And so I probably over specified it a bit using a Gantt chart,
link |
but the fact that you can do it,
link |
it's what professional kitchens roughly do.
link |
They just don't call it a Gantt chart,
link |
or at least I don't think they do.
link |
Anyway, I was telling a story about Thanksgiving.
link |
So here's the thing.
link |
I'm sitting down, we have the meal,
link |
and then I got to know Ray Dalio fairly well
link |
over maybe the last year of Instagram.
link |
And one thing that he kept saying was like,
link |
feedback is really hard to get honestly from people.
link |
And I sat down after dinner, I said,
link |
guys, I want feedback.
link |
What was good and what was bad?
link |
And what's funny is like,
link |
literally everyone just said everything was great.
link |
And I like personally knew I had screwed up
link |
a handful of things, but no one would say it.
link |
And can you imagine now not something as high stakes
link |
as Thanksgiving dinner, okay?
link |
Thanksgiving dinner, it's not that high stakes.
link |
But you're trying to build a product
link |
and everyone knows you left your job for it
link |
and you're trying to build it out
link |
and you're trying to make something wonderful
link |
and it's yours, right?
link |
Now try asking for feedback
link |
and know that you're giving this to your friends
link |
People have trouble giving hard feedback.
link |
People have trouble saying, I don't like this
link |
or this isn't great, or this is how it's failed me.
link |
In fact, you usually have two classes of people.
link |
People who just won't say bad things,
link |
you can literally say to them,
link |
please tell me what you hate most about this
link |
and they won't do it.
link |
They'll try, but they won't.
link |
And then the other class of people
link |
are just negative period about everything
link |
and it's hard to parse out like what is true and what isn't.
link |
So my rule of thumb with this is
link |
you should always ask people,
link |
but at the end of the day, it's amazing what data
link |
And that's why with whatever project I work on, even now,
link |
collecting data from the beginning on usage patterns,
link |
so engagement, how many days of the week do they use it?
link |
How many, I don't know if we were to go back to Instagram,
link |
how many impressions per day, right?
link |
Is that shrinking?
link |
And don't be like overly scientific about it, right?
link |
Cause maybe you have 50 beta users or something.
link |
But what's fascinating is that data doesn't lie.
link |
People are very defensive about their time.
link |
They'll say, oh, I'm so busy,
link |
I'm sorry I didn't get to use the app.
link |
Like I'm just, you know,
link |
but I don't know you were posting on Instagram
link |
So I don't know at the end of the day,
link |
like at Facebook there was, you know,
link |
before time spent became kind of this loaded term there.
link |
The idea that people's currency in their lives is time.
link |
And they only have a certain amount of time to give things,
link |
whether it's friends or family or apps or TV shows
link |
or whatever, there's no way of inventing more of it,
link |
at least not that I know of.
link |
If they don't use it, it's because it's not great.
link |
So the moral of the story is you can ask all you want,
link |
but you just have to look at the data.
link |
And data doesn't lie, right?
link |
I mean, there's metrics, there's data can obscure
link |
the key insight if you're not careful.
link |
So time spent in the app, that's one.
link |
There's so many metrics you can put at this
link |
and they will give you totally different insights,
link |
especially when you're trying to create something
link |
that doesn't obviously exist yet.
link |
So, you know, measuring maybe why you left the app
link |
or measuring special moments of happiness
link |
that will make sure you return to the app
link |
or moments of happiness that are long lasting
link |
versus like dopamine short term, all of those things.
link |
But I think I suppose in the beginning,
link |
you can just get away with just asking the question,
link |
which features are used a lot?
link |
Let's do more of that.
link |
And how hard was the decision?
link |
And I mean, maybe you can tell me what Instagram
link |
looked in the beginning,
link |
but how hard was it to make pictures
link |
of the first class citizen?
link |
That's a revolutionary idea.
link |
Like at whatever point Instagram became this feed of photos,
link |
that's quite brilliant.
link |
Plus, I also don't know when this happened,
link |
but they're all shaped the same.
link |
I have to tell you why, that's the interesting part.
link |
So a couple of things.
link |
One is data, like you're right.
link |
You can overinterpret data.
link |
Like imagine trying to fly a plane by staring at,
link |
I don't know, a single metric like airspeed.
link |
You don't know if you're going up or down.
link |
I mean, it correlates with up or down,
link |
but you don't actually know.
link |
It will never help you land the plane.
link |
So don't stare at one metric.
link |
Like it turns out you have to synthesize a bunch of metrics
link |
to know where to go.
link |
But it doesn't lie.
link |
Like if your airspeed is zero,
link |
unless it's not working, right?
link |
If it's zero, you're probably gonna fall out of the sky.
link |
So generally you look around and you have the scan going.
link |
And you're just asking yourself,
link |
is this working or is this not working?
link |
But people have trouble explaining how they actually feel.
link |
So just, it's about synthesizing both of them.
link |
So then Instagram, right?
link |
We were talking about revolutionary moment
link |
where the feed became square photos basically.
link |
And photos first and then square photos.
link |
It was clear to me that the biggest,
link |
so I believe the biggest companies are founded
link |
when enormous technical shifts happen.
link |
And the biggest technical shift that happened
link |
right before Instagram was founded
link |
was the advent of a phone that didn't suck.
link |
The iPhone, right?
link |
Like in retrospect, we're like, oh my God,
link |
the first iPhone that almost had,
link |
like it wasn't that good.
link |
But compared to everything else at the time,
link |
the first phone that had an incredible camera
link |
that could like do as well as the point and shoot
link |
you might carry around was the iPhone 4.
link |
And that was right when Instagram launched.
link |
And we looked around and we said,
link |
Because everyone has a camera in their pocket.
link |
And it was so clear to me that the world
link |
of social networks before it was based in the desktop
link |
and sitting there and having a link you could share, right?
link |
And that wasn't gonna be the case.
link |
So the question is what would you share
link |
if you were out and about in the world?
link |
If not only did you have a camera that fit in your pocket,
link |
but by the way, that camera had a network attached to it
link |
that allowed you to share instantly.
link |
That seemed revolutionary.
link |
And a bunch of people saw it at the same time.
link |
It wasn't just Instagram.
link |
There were a bunch of competitors.
link |
The thing we did, I think was not only,
link |
well, we focused on two things.
link |
So we wrote down those things, we circled photos
link |
and we said, I think we should invest in this.
link |
But then we said, what sucks about photos?
link |
One, they look like crap, right?
link |
They just, at least back then.
link |
Now my phone takes pretty great photos, right?
link |
Back then they were blurry, not so great, compressed, right?
link |
Two, it was really slow, like really slow to upload a photo.
link |
And I'll tell a fun story about that
link |
and explain to you why they're all the same size
link |
and square as well.
link |
And three, man, if you wanted to share a photo
link |
on different networks, you had to go to each
link |
of the individual apps and select all of them
link |
and upload individually.
link |
And so we were like, all right, those are the pain points.
link |
We're gonna focus on that.
link |
So one, instead of, because they weren't beautiful,
link |
we were like, why don't we lean into the fact
link |
that they're not beautiful?
link |
And I remember studying in Florence,
link |
my photography teacher gave me this Holga camera
link |
and I'm not sure everyone knows what a Holga camera is,
link |
but they're these old school plastic cameras.
link |
I think they're produced in China at the time.
link |
And I wanna say the original ones were like
link |
from the 70s or the 80s or something.
link |
They're supposed to be like $3 cameras for the every person.
link |
They took nice medium format films, large negatives,
link |
but they kind of blurred the light
link |
and they kind of like light leaked into the side.
link |
And there was this whole resurgence where people looked
link |
at that and said, oh my God, this is a style, right?
link |
And I remember using that in Florence and just saying,
link |
well, why don't we just like lean into the fact
link |
that these photos suck and make them suck more,
link |
but in an artistic way.
link |
And it turns out that had product market fit.
link |
People really liked that.
link |
They were willing to share their not so great photos
link |
if they looked not so great on purpose, okay?
link |
That's where the filters come into picture.
link |
So computational modification of photos
link |
to make them look extra crappy to where it becomes art.
link |
And I mean, add light leaks, add like an overlay filter,
link |
make them more contrasty than they should be.
link |
The first filter we ever produced was called X Pro 2.
link |
And I designed it while I was
link |
in this small little bed and breakfast room
link |
in Todos Santos, Mexico.
link |
I was trying to take a break from the bourbon days.
link |
And I remember saying to my co founder,
link |
I just need like a week to reset.
link |
And that was on that trip worked on the first filter
link |
because I said, you know, I think I can do this.
link |
And I literally iterated one by one over the RGB values
link |
in the array that was the photo and just slightly shifted.
link |
Basically there was a function of R, function of G,
link |
function of B that just shifted them slightly.
link |
It was in rocket science.
link |
And it turns out that actually
link |
made your photo look pretty cool.
link |
It just mapped from one color space to another color space.
link |
It was simple, but it was really slow.
link |
I mean, if you applied a filter,
link |
I think it used to take two or three seconds to render.
link |
Only eventually would I figure out how to do it on the GPU.
link |
And I'm not even sure it was a GPU, but it was using OpenGL.
link |
But anyway, I would eventually figure that out
link |
and then it would be instant, but it used to be really slow.
link |
By the way, anyone who's watching or listening,
link |
it's amazing what you can get away with in a startup
link |
as long as the product outcome is right for the user.
link |
Like you can be slow.
link |
You can be terrible.
link |
You can be, as long as you have product market fit,
link |
people will put up with a lot.
link |
And then the question is just about compressing,
link |
making it more performant over time
link |
so that they get that product market fit instantly.
link |
So fascinating because there's some things
link |
where those three seconds would make or break the app,
link |
but some things you're saying not.
link |
It's hard to know when, it's the problem Spotify solved
link |
for making streaming like work.
link |
And like delays in listening to music is a huge negative,
link |
even like slight delays.
link |
But here you're saying, I mean,
link |
how do you know when those three seconds are okay
link |
or are you just gonna have to try it out?
link |
Because to me, my intuition would be
link |
those three seconds would kill the app.
link |
Like I would try to do the OpenGL thing.
link |
Right, so I wish I were that smart at the time.
link |
I wasn't, I just knew how to do what I knew how to do.
link |
And I decided, okay, like,
link |
why don't I just iterate over the values and change them?
link |
And what's interesting is that
link |
compared to the alternatives, no one else used OpenGL.
link |
So everyone else was doing it the dumb way.
link |
And in fact, they were doing it at a high resolution.
link |
Now comes in the small resolution
link |
that we'll talk about in a second.
link |
By choosing 512 pixels by 512 pixels,
link |
which I believe it was at the time,
link |
we iterated over a lot fewer pixels than our competitors
link |
who were trying to do these enormous output like images.
link |
So instead of taking 20 seconds,
link |
I mean, three seconds feels pretty good, right?
link |
So on a relative basis, we were winning like a lot.
link |
Okay, so that's answer number one.
link |
Answer number two is we actually focused
link |
on latency in the right places.
link |
So we did this really wonderful thing when you uploaded.
link |
So the way it would work is you'd take your phone,
link |
you'd take the photo and then you'd go to the,
link |
you'd go to the edit screen where you would caption it.
link |
And on that caption screen, you'd start typing
link |
and you'd think, okay, like what's a clever caption?
link |
And I said to Mike, hey, when I worked on the Gmail team,
link |
you know what they did?
link |
When you typed in your username or your email address,
link |
even before you've entered in your password,
link |
like the probability once you enter in your username
link |
that you're going to actually sign in is extremely high.
link |
So why not just start loading your account
link |
in the background?
link |
Not like sending it down to the desktop,
link |
that would be a security issue,
link |
but like loaded into memory on the server,
link |
like get it ready, prepare it.
link |
I always thought that was so fascinating and unintuitive.
link |
And I was like, Mike, why don't we just do that?
link |
But like, we'll just upload the photo
link |
and like assume you're going to upload the photo.
link |
And if you don't, forget about it, we'll delete it, right?
link |
So what ended up happening was people would caption
link |
their photo, they'd press done or upload
link |
and you'd see this little progress bar just go foop.
link |
It was lightning fast, okay?
link |
We were no faster than anyone else at the time,
link |
but by choosing 512 by 512 and doing it in the background,
link |
it almost guaranteed that it was done
link |
by the time you captioned.
link |
And everyone when they used it was like,
link |
how the hell is this thing so fast?
link |
But we were slow, we just hid the slowness.
link |
It wasn't like, these things are just like,
link |
it's a shell game, you're just hiding the latency.
link |
That mattered to people like a lot.
link |
And I think that, so you were willing to put up
link |
with a slow filter if it meant
link |
you could share it immediately.
link |
And of course we added sharing options
link |
which let you distribute it really quickly,
link |
that was the third part.
link |
So latency matters, but relative to what?
link |
And then there's some like tricks,
link |
you get around to just hiding the latency.
link |
Like I don't know if Spotify starts downloading
link |
the next song eagerly, I'm assuming they do.
link |
There are a bunch of ideas here that are not rocket science
link |
And all of that was stuff you were explicitly
link |
having a discussion about, like those designs
link |
and you were having like arguments, discussions.
link |
I'm not sure it was arguments, I mean,
link |
I'm not sure if you've met my co founder Mike,
link |
but he's a pretty nice guy and he's very reasonable.
link |
And we both just saw eye to eye and we're like,
link |
yeah, it's like, make this fast or at least seem fast,
link |
Honestly, I think the most contentious thing
link |
and he would say this too initially,
link |
was I was on an iPhone 3G, so like the not so fast one.
link |
And he had a brand new iPhone 4, that was cheap.
link |
And his feed loaded super smoothly,
link |
like when he would scroll from photo to photo,
link |
buttery smooth, right?
link |
But on my phone, every time you got to a new photo,
link |
it was like, kachunk, kachunk, allocate memory,
link |
like all this stuff, right?
link |
I was like, Mike, that's unacceptable.
link |
He's like, oh, come on, man, just like upgrade your phone.
link |
Basically, he didn't actually say that,
link |
he was nicer than that.
link |
But I could tell he wished like,
link |
I would just stop being cheap and just get a new phone.
link |
But what's funny is we actually sat there working
link |
on that little detail for a few days before launch.
link |
And that polished experience,
link |
plus the fact that uploading seemed fast
link |
for all these people who didn't have nice phones,
link |
I think meant a lot because far too often,
link |
you see teams focus not on performance,
link |
they focus on what's the cool computer science problem
link |
they can solve, right?
link |
Can we scale this thing to a billion users
link |
and they've got like a hundred, right?
link |
You talked about loss function,
link |
so I want to come back to that.
link |
Like the loss function is like,
link |
do you provide a great, happy, magical,
link |
whatever experience for the consumer?
link |
And listen, if it happens to involve something complex
link |
and technical, then great.
link |
But it turns out, I think most of the time,
link |
those experiences are just sitting there waiting to be built
link |
with like not that complex solutions.
link |
But everyone is just like so stuck in their own head
link |
that they have to overengineer everything
link |
and then they forget about the easy stuff.
link |
I mean, also, maybe to flip the loss function there is,
link |
you're trying to minimize the number of times
link |
there's unpleasant experience, right?
link |
Like the one you mentioned where when you go
link |
to the next photo, it freezes for a little bit.
link |
So it's almost, as opposed to maximizing pleasure,
link |
it's probably easier to minimize the number of like,
link |
And as we all know, you just make the pleasure negative
link |
and then minimize everything, so.
link |
We're mapping this all back to neural networks.
link |
But actually, can I say one thing on that,
link |
which is I don't know a lot about machine learning,
link |
but I feel like I've tried studying a bunch.
link |
That whole idea of reinforcement learning
link |
and planning out more than the greedy single experience,
link |
I think is the closest you can get
link |
to like ideal product design thinking,
link |
where you're not saying,
link |
hey, like, can we have a great experience just this one time?
link |
But like, what is the right way to onboard someone?
link |
What series of experiences correlate most with them
link |
hanging on long term, right?
link |
So not just saying, oh, did the photo load slowly
link |
a couple of times, or did they get a great photo
link |
at the top of their feed?
link |
But like, what are the things that are gonna make
link |
this person come back over the next week,
link |
over the next month?
link |
And as a product designer asking yourself,
link |
okay, I wanna optimize, not just minimize bad experiences
link |
in the short run, but like,
link |
how do I get someone to engage over the next month?
link |
And I'm not gonna claim at all that I thought that way
link |
at all at the time, because I certainly didn't.
link |
But if I were going back and giving myself any advice,
link |
it would be thinking, what are those second order effects
link |
that you can create?
link |
And it turns out having your friends on the service
link |
is an enormous win.
link |
So starting with a very small group of people
link |
that produce content that you wanted to see, which we did,
link |
we seeded the community very well, I think.
link |
Ended up mattering, and so.
link |
Yeah, you said that community is one
link |
of the most important things.
link |
So it's from a metrics perspective,
link |
from maybe a philosophy perspective,
link |
building a certain kind of community within the app.
link |
See, I wasn't sure what exactly you meant by that
link |
when I first heard you say that.
link |
Maybe you can elaborate, but as I understand now,
link |
it can literally mean get your friends onto the app.
link |
Yeah, think of it this way.
link |
You can build an amazing restaurant or bar or whatever,
link |
right, but if you show up and you're the only one there,
link |
is it like, does it matter how good the food is?
link |
The drinks, whatever?
link |
No, these are inherently social experiences
link |
that we were working on.
link |
So the idea of having people there,
link |
like you needed to have that,
link |
otherwise it was just a filter out.
link |
But by the way, part of the genius,
link |
I'm gonna say genius, even though it wasn't really genius,
link |
was starting to be marauding as a filter app was awesome.
link |
The fact that you could,
link |
so we talk about single player mode a lot,
link |
which is like, can you play the game alone?
link |
And Instagram, you could totally play alone.
link |
You could filter your photos,
link |
and a lot of people would tell me,
link |
I didn't even realize that this thing was a social network
link |
until my friend showed up.
link |
It totally worked as a single player game.
link |
And then when your friends showed up,
link |
all of a sudden it was like,
link |
oh, not only was this great alone,
link |
but now I actually have this trove of photos
link |
that people can look at and start liking,
link |
and then I can like theirs.
link |
And so it was this bootstrap method
link |
of how do you make the thing not suck
link |
when the restaurant is empty?
link |
Yeah, but the thing is, when you say friends,
link |
we're not necessarily referring to friends
link |
in the physical space.
link |
So you're not bringing your physical friends with you.
link |
You're also making new friends.
link |
So you're finding new community.
link |
So it's not immediately obvious to me
link |
that it's almost like building any kind of community.
link |
And what we learned very early on
link |
was what made Instagram special
link |
and the reason why you would sign up for it
link |
versus say, just sit on Facebook
link |
and look at your friends photos.
link |
Of course we were live,
link |
and of course it was interesting
link |
to see what your friends were doing now.
link |
But the fact that you could connect with people
link |
who like took really beautiful photos in a certain style
link |
all around the world, whether they were travelers,
link |
it was the beginning of the influencer economy.
link |
It was these people
link |
who became professional Instagrammers way back when.
link |
But they took these amazing photos
link |
and some of them were photographers professionally.
link |
And all of a sudden you had this moment in the day
link |
when you could open up this app
link |
and sure you could see what your friends were doing,
link |
but also it was like, oh my God,
link |
that's a beautiful waterfall or oh my God,
link |
I didn't realize there was that corner of England
link |
or like really cool stuff.
link |
And the beauty about Instagram early on
link |
was that it was international by default.
link |
You didn't have to speak English to use it, right?
link |
You could just look at the photos, works great.
link |
We did translate, we had some pretty bad translations,
link |
but we did translate the app.
link |
And even if our translations were pretty poor,
link |
the idea that you could just connect with other people
link |
through their images was pretty powerful.
link |
So how much technical difficulties
link |
there with the programming?
link |
Like what programming language you were talking about?
link |
Zero, like maybe it was hard for us,
link |
but I mean, there was nothing.
link |
The only thing that was complex about Instagram
link |
at the beginning technically was making it scale.
link |
We were just plain old Objective C for the client.
link |
So it was iPhone only at first?
link |
As an Android person, I'm deeply offended, but go ahead.
link |
Android's getting a lot better, right?
link |
I take it back, you're right.
link |
If I were to do something today,
link |
I think it would be very different
link |
in terms of launch strategy, right?
link |
Android's enormous too.
link |
But anyway, back to that moment,
link |
it was Objective C and then we were Python based,
link |
which is just like, this is before Python was really cool.
link |
Like now it's cool
link |
because it's all these machine learning libraries
link |
like support Python and right.
link |
Now it's super, now it's like cool to be Python.
link |
Back then it was like, oh, Google uses Python.
link |
Like maybe you should use Python.
link |
Like I had worked at a small startup
link |
of some ex Googlers that used Python.
link |
So we used it and we used a framework called Django,
link |
still exists and people use for basically the backend.
link |
And then you threw a couple interesting things in there.
link |
I mean, we used Postgres, which was kind of fun.
link |
It was a little bit like Hipster database
link |
at the time, right?
link |
MySQL, like everyone used MySQL.
link |
So like using Postgres was like an interesting decision.
link |
But we used it because it had a bunch of geo features
link |
built in because we thought we were gonna be
link |
a check and app, remember?
link |
It's also super cool now.
link |
So you were into Python before it was cool
link |
and you were into Postgres before it was cool.
link |
Yeah, we were basically like,
link |
not only Hipster photo company,
link |
Hipster tech company, right?
link |
We also adopted Redis early and like loved it.
link |
I mean, it solves so many problems for us
link |
and turns out that's still pretty cool.
link |
But the programming was very easy.
link |
It was like, sign up a user, have a feed.
link |
There was nothing, no machine learning at all, zero.
link |
Can you get some context?
link |
How many users at each of these stages?
link |
Are we talking about a hundred users, a thousand users?
link |
So the stage I just described,
link |
I mean that technical stack lasted
link |
through probably 50 million users.
link |
I mean, seriously, like you can get away with a lot
link |
with a pretty basic stack.
link |
Like I think a lot of startups try to overengineer
link |
their solutions from the beginning to like really scale
link |
and you can get away with a lot.
link |
That being said, most of the first two years of Instagram
link |
was literally just trying to make that stack scale.
link |
And it wasn't, it was not a Python problem.
link |
It was like literally just like, where do we put the data?
link |
Like it's all coming in too fast.
link |
Like how do we store it?
link |
How do we make sure to be up?
link |
How do we like, how do we make sure we're
link |
on the right size boxes that they have enough memory?
link |
Those were the issues, but.
link |
Can you speak to the choices you make at that stage
link |
when you're growing so quickly?
link |
Do you use something like somebody else's
link |
computer infrastructure or do you build in house?
link |
I'm only laughing because we, when we launched
link |
we had a single computer that we had rented
link |
in some color space in LA.
link |
I don't even remember what it was called.
link |
Cause I thought that's what you did.
link |
When I worked at a company called Odio that became Twitter.
link |
I remember visiting our space in San Francisco.
link |
You walked in, you had to wear the ear things.
link |
It was cold and fans everywhere, right?
link |
And we had to, you know, plug one out, replace one.
link |
And I was the intern, so I just like held things.
link |
But I thought to myself, oh, this is how it goes.
link |
And then I remember being in a VC's office.
link |
I think it was a benchmarks office.
link |
And I think we ran into another entrepreneur
link |
and they were like, oh, how are things going?
link |
We're like, ah, you know, try to scale this thing.
link |
And they were like, well, I mean
link |
can't you just add more instances?
link |
And I was like, what do you mean?
link |
And they're like instances on Amazon.
link |
I was like, what are those?
link |
And it was this moment where we realized how deep
link |
in it we were because we had no idea that AWS existed
link |
nor should we be using it.
link |
Anyway, that night we went back to the office
link |
and we got on AWS, but we did this really dumb thing.
link |
We're so sorry to people listening.
link |
But we brought up an instance, which was our database.
link |
It's gonna be a replacement for our database.
link |
But we had it talking over the public internet
link |
to our little box in LA that was our app server.
link |
That's how sophisticated we were.
link |
And obviously that was very, very slow.
link |
Didn't work at all.
link |
I mean, it worked, but didn't work.
link |
Only like later that night did we realize
link |
we had to have it all together.
link |
But at least like if you're listening right now
link |
and you're thinking, you know, I have no chance.
link |
I'm gonna start to start, but I have no chance.
link |
We did it and we made a bunch
link |
of really dumb mistakes initially.
link |
I think the question is how quickly do you learn
link |
that you're making a mistake?
link |
And do you do the right thing immediately right after?
link |
So you didn't pay for those mistakes by failure.
link |
So yeah, how quickly did you fix it?
link |
I guess there's a lot of ways to sneak up to this question
link |
of how the hell do you scale the thing?
link |
Other startups, if you have an idea,
link |
how do you scale the thing?
link |
Is it just AWS and you try to write the kind of code
link |
that's easy to spread across a large number of instances,
link |
and then the rest is just put money into it?
link |
Basically, I would say a couple of things.
link |
First off, don't even ask the question,
link |
just find product market fit, duct tape it together, right?
link |
Like if you have to.
link |
I think there's a big caveat here, which I want to get to,
link |
but generally all that matters is product market fit.
link |
That's all that matters.
link |
If people like your product,
link |
do not worry about when 50,000 people use your product
link |
because you will be happy that you have that problem
link |
when you get there.
link |
I actually can't name many startups
link |
where they go from nothing to something overnight
link |
and they can't figure out how to scale it.
link |
There are some, but I think nowadays,
link |
it's a, when I say a solved problem,
link |
like there are ways of solving it.
link |
The base case is typically that startups
link |
worry way too much about scaling way too early
link |
and forget that they actually have to make something
link |
That's the default mistake case.
link |
But what I'll say is once you start scaling,
link |
I mean, hiring quickly,
link |
people who have seen the game before
link |
and just know how to do it,
link |
it becomes a bit of like, yeah,
link |
just throw instances of the problem, right?
link |
But the last thing I'll say on this
link |
that I think did save us,
link |
we were pretty rigorous about writing tests
link |
from the beginning.
link |
That helped us move very, very quickly
link |
when we wanted to rewrite parts of the product
link |
and know that we weren't breaking something else.
link |
Tests are one of those things where it's like,
link |
you go slow to go fast.
link |
And they suck when you have to write them
link |
because you have to figure it out.
link |
And there are always those ones that break
link |
when you don't want them to break and they're annoying
link |
and it feels like you spent all this time.
link |
But looking back, I think that like longterm optimal,
link |
even with a team of four,
link |
it allowed us to move very, very quickly
link |
because anyone could touch any part of the product
link |
and know that they weren't gonna bring down the site,
link |
or at least in general.
link |
At which point do you know product market fit?
link |
How many users would you say?
link |
Is it all it takes is like 10 people?
link |
Or is it a thousand?
link |
I don't think it is generally
link |
a question of absolute numbers.
link |
I think it's a question of cohorts
link |
and I think it's a question of trends.
link |
So, you know, it depends how big your business is trying
link |
But if I were signing up a thousand people a week
link |
and they all retain,
link |
like the retention curves for those cohorts looked good,
link |
healthy, and even like,
link |
as you started getting more people on the service,
link |
maybe those earlier cohorts started curving up again
link |
because now there are network effects
link |
and their friends are on the service
link |
or totally depends what type of business you're in,
link |
but I'm talking purely social, right?
link |
I don't think it's an absolute number.
link |
I guess you could call it a marginal number.
link |
So I spent a lot of time when I work with startups
link |
asking them like, okay,
link |
have you looked at that cohort versus this cohort,
link |
whether it's your clients
link |
or whether it's people signing up for the service?
link |
But a lot of people think you just have to hit some mark,
link |
like 10,000 people or 50,000 people,
link |
but really seven ish billion people in the world.
link |
Most people forever will not know about your product.
link |
There are always more people out there to sign up.
link |
It's just a question of how you turn on the spigot, so.
link |
At that stage, early stage yourself,
link |
but also by way of advice,
link |
should you worry about money at all?
link |
How this thing's gonna make money?
link |
Or do you just try to find product market fit
link |
and get a lot of users to enjoy using your thing?
link |
I think it totally depends.
link |
And that's an unsatisfying answer.
link |
I was talking with a friend today who,
link |
he was one of our earlier investors and he was saying,
link |
hey, like, have you been doing any angel investing lately?
link |
I said, not really.
link |
I'm just like focused on what I wanna do next.
link |
And he said, the number of financings have just gone bonkers.
link |
Like people are throwing money everywhere right now.
link |
And I think the question is,
link |
do you have an inkling of how you're gonna make money?
link |
Or are you really just like waving your hands?
link |
I would not like to be an entrepreneur in the position of,
link |
well, I have no idea how this will eventually make money.
link |
If you are in an area,
link |
like let's say you wanted to start a social network, right?
link |
Not saying this is a good idea, but if you did,
link |
there are only a handful of ways they've made money
link |
and really only one way they've made money in the past
link |
So, if you have a service that's amenable to that
link |
and then I wouldn't worry too much about that
link |
because if you get to the scale,
link |
you can hire some smart people and figure that out.
link |
I do think that is really healthy for a lot of startups
link |
these days, especially the ones doing
link |
like enterprise software, slacks of the world, et cetera,
link |
to be worried about money from the beginning,
link |
but mostly as a way of winning over clients
link |
and having stickiness.
link |
Of course you need to be worried about money,
link |
but I'm gonna also say this again,
link |
which is it's like longterm profitability.
link |
If you have a roadmap to that, then that's great.
link |
But if you're just like, I don't know, maybe never,
link |
like we're working on this metaverse thing,
link |
I think maybe someday, I don't know.
link |
Like that seems harder to me.
link |
So you have to be as big as Facebook
link |
to like finance that bet, right?
link |
Do you think it's possible, you said,
link |
you're not saying it's necessarily a good idea
link |
to launch a social network.
link |
Do you think it's possible today,
link |
maybe you can put yourself in those shoes,
link |
to launch a social network that achieves the scale
link |
of a Facebook or a Twitter or an Instagram,
link |
and maybe even greater scale?
link |
Asking for a friend.
link |
Yeah, if I knew, I'd probably be doing it right now
link |
and not sitting here.
link |
So, I mean, there's a lot of ways to ask this question.
link |
One is create a totally new product market fit,
link |
create a new market, create something like Instagram did,
link |
which is like create something kind of new,
link |
or literally out compete Facebook at its own thing,
link |
or I'll compete Twitter at its own thing.
link |
The only way to compete now,
link |
if you wanna build a large social network
link |
is to look for the cracks, look for the openings.
link |
No one competed, I mean,
link |
no one competed with the core business of Google.
link |
No one competed with the core business of Microsoft.
link |
You don't go at the big guys
link |
doing exactly what they're doing.
link |
Instagram didn't win, quote unquote,
link |
because it tried to be a visual Twitter.
link |
Like we spotted things that either Twitter
link |
wasn't going to do or refuse to do,
link |
images and feed for the longest time, right?
link |
Or that Facebook wasn't doing or not paying attention to
link |
because they were mostly desktop at the time
link |
and we were purely mobile, purely visual.
link |
Often there are opportunities sitting there.
link |
You have to figure out like,
link |
I think like there's a strategy book,
link |
I can't remember the name,
link |
but talk about moats and just like the best place to play
link |
is where your competitor literally can't pivot
link |
because structurally they're set up not to be there.
link |
And that's where you win.
link |
And what's fascinating is like,
link |
do you know how many people were like,
link |
images, Facebook does that, Twitter does that.
link |
I mean, how wrong were they, really wrong?
link |
And these are some of the smartest people
link |
in Silicon Valley, right?
link |
But now Instagram exists for a while.
link |
How is it that Snapchat could then exist?
link |
It makes no sense.
link |
Like plenty of people would say,
link |
well, there's Facebook, no images.
link |
Okay, okay, Instagram, I'll give you that one.
link |
But wait, now another image based social network
link |
is gonna get really big.
link |
And then TikTok comes along.
link |
Like the prior, so you asked me, is it possible?
link |
The only reason I'm answering yes
link |
is because my prior is that it's happened once every,
link |
I don't know, three, four or five years consistently.
link |
And I can't imagine there's anything structurally
link |
that would change that.
link |
So that's why I answer that way.
link |
Not because I know how, I just,
link |
when you see a pattern, you see a pattern
link |
and there's no reason to believe that's gonna stop.
link |
And it's subtle too, because like you said,
link |
Snapchat and TikTok,
link |
they're all doing the same space of things,
link |
but there's something fundamentally different
link |
about like a three second video and a five second video
link |
and a 15 second video and a one minute video
link |
and a one hour video, like fundamentally different.
link |
Fundamentally different.
link |
I mean, I think one of the reasons Snapchat exists
link |
is because Instagram was so focused on posting great,
link |
beautiful manicured versions of yourself throughout time.
link |
And there was this enormous demand of like,
link |
hey, I really like this behavior.
link |
I love using Instagram, but man,
link |
I just like wish I could share something going on in my day.
link |
Do I really have to put it on my profile?
link |
Do I really have to make it last forever?
link |
Do I really, and that opened up a door,
link |
it created a market, right?
link |
And then what's fascinating is Instagram had an explore page
link |
for the longest time and it was image driven, right?
link |
But there's absolutely a behavior where you open up Instagram
link |
and you sit on the explore page all day.
link |
That is effectively TikTok,
link |
but obviously focused on videos.
link |
And it's not like you could just put the explore page
link |
in TikTok form and it works.
link |
It had to be video, it had to have music.
link |
These are the hard parts about product development
link |
that are very hard to predict,
link |
but they're all versions of the same thing
link |
with varying, like if you line them up
link |
in a bunch of dimensions, they're just like kind of on,
link |
they're different values of the same dimensions,
link |
which is like, I guess, easy to say in retrospect.
link |
But like, if I were an entrepreneur going after that area,
link |
I'd ask myself like, where's the opening?
link |
What needs to exist because TikTok exists now?
link |
So I wonder how much things that don't yet exist
link |
and can exist is in the space of algorithms,
link |
in the space of recommender systems.
link |
So in the space of how the feed is generated.
link |
So we kind of talk about the actual elements
link |
of the content, that's what we've been talking,
link |
the difference between photos,
link |
between short videos, longer videos.
link |
I wonder how much disruption is possible
link |
in the way the algorithms work.
link |
Because a lot of the criticism towards social media
link |
is in the way the algorithms work currently.
link |
And it feels like, first of all,
link |
talking about product market fit,
link |
there's certainly a hunger for social media algorithms
link |
that do something different.
link |
I don't think anyone, everyone said complaining,
link |
this is hurting me and this is hurting society,
link |
but I keep doing it because I'm addicted to it.
link |
And they say, we want something different,
link |
but we don't know what.
link |
It feels like just different.
link |
It feels like there's a hunger for that,
link |
but that's in the space of algorithms.
link |
I wonder if it's possible to disrupt in that space.
link |
Absolutely, I have this thesis that the worst part
link |
about social networks is that they're, is the people.
link |
It's a line that sounds funny, right?
link |
Because like, that's why you call it a social network.
link |
But what does social networks actually do for you?
link |
Like just think, like imagine you were an alien
link |
and you landed and someone says,
link |
hey, there's this site, it's a social network.
link |
We're not gonna tell you what it is,
link |
but just what does it do?
link |
And you have to explain it to them.
link |
It does two things.
link |
One is that people you know and have social ties with
link |
distribute updates through whether it's photos or videos
link |
about their lives so that you don't have to physically
link |
be with them, but you can keep in touch with them.
link |
That's one, that's like a big part of Instagram.
link |
That's a big part of Snap.
link |
It is not part of TikTok at all.
link |
So there's another big part, which is there's all this
link |
content out in the world that's entertaining,
link |
whether you wanna watch it or you wanna read it.
link |
And matchmaking between content that exists in the world
link |
and people that want that content
link |
turns out to be like a really big business, right?
link |
Search and discovery, which you?
link |
Search and discovery, but my point is it could be video,
link |
it could be text, it could be websites, it could be,
link |
I mean, think back to like dig, right?
link |
Or stumble upon or, right?
link |
But like, what did those do?
link |
Like they basically distributed interesting content
link |
I think the most interesting part or the future
link |
of social networks is going to be making them less social
link |
because I think people are part of the root cause
link |
So for instance, often in recommender systems,
link |
we talk about two stages.
link |
There's a candidate generation step, which is just like
link |
of our vast trove of stuff that you might wanna see,
link |
what small subset should we pick for you, okay?
link |
Typically that is grabbed from things
link |
your friends have shared, right?
link |
Then there's a ranking step which says, okay,
link |
now given these 100, 200 things depends on the network,
link |
Let's like be really good about ranking them
link |
and generally rank the things up higher
link |
that get the most engagement, right?
link |
So what's the problem with that?
link |
Step one is we've limited everything you could possibly see
link |
to things that your friends have chosen to share
link |
or maybe not friends, but influencers.
link |
What things do people generally want to share?
link |
They wanna share things that are gonna get likes,
link |
that are gonna show up broadly.
link |
So they tend to be more emotionally driven.
link |
They tend to be more risque or whatever.
link |
So why do we have this problem?
link |
It's because we show people things people have decided
link |
to share and those things self select to being the things
link |
that are most divisive.
link |
So how do you fix that?
link |
Well, what if you just imagine for a second
link |
that why do you have to grab things
link |
from things your friends have shared?
link |
Why not just like grab things?
link |
That's really fascinating to me.
link |
And that's something I've been thinking a lot about.
link |
And just like, why is it that when you log onto Twitter,
link |
you're just sitting there looking at things from accounts
link |
that you've followed for whatever reason?
link |
And TikTok I think has done a wonderful job here,
link |
which is like, you can literally be anyone.
link |
And if you produce something fascinating, it'll go viral.
link |
But like, you don't have to be someone that anyone knows.
link |
You don't have to have built up a giant following.
link |
You don't have to have paid for followers.
link |
You don't have to try to maintain those followers.
link |
You literally just have to produce something interesting.
link |
That is I think the future of social networking.
link |
That's the direction things will head.
link |
And I think what you'll find is it's far less
link |
about people manipulating distribution
link |
and far more about what is like, is this content good?
link |
And good is obviously a vague definition
link |
that we could spend hours on.
link |
But different networks I think will decide
link |
different value functions to decide what is good
link |
and what isn't good.
link |
And I think that's a fascinating direction.
link |
So that's almost like creating an internet.
link |
I mean, that's what Google did for web pages
link |
that did page rank search.
link |
So it's discovery, you don't follow anybody on Google
link |
when you use a search engine.
link |
You just discover web pages.
link |
And so what TikTok does is saying,
link |
let's start from scratch.
link |
Let's like start a new internet
link |
and have people discover stuff on that new internet
link |
within a particular kind of pool of people.
link |
But what's so fascinating about this
link |
is like the field of information retrieval.
link |
Like I always talked about as I was studying this stuff,
link |
it was used the word query and document.
link |
So I was like, why are they saying query and documents?
link |
Like they're literally imagine,
link |
like if you just stop thinking about query
link |
as like literally a search query
link |
and a query could be a person.
link |
I mean, a lot of the way,
link |
I'm not gonna claim to know how Instagram or Facebook
link |
machine learning works today,
link |
but if you want to find a match for a query,
link |
the query is actually the attributes of the person,
link |
their age, their gender, where they're from,
link |
maybe some kind of summarization of their interests.
link |
And that's a query, right?
link |
And that matches against documents.
link |
And by the way, documents don't have to be texts.
link |
They can be videos, however long.
link |
I don't know what the limit is on TikTok these days.
link |
They keep changing it.
link |
My point is just, you've got a query,
link |
which is someone in search of something
link |
that they want to match and you've got the document
link |
and it doesn't have to be text.
link |
It could be anything.
link |
And how do you match make?
link |
And that's one of these like,
link |
I mean, I've spent a lot of time thinking about this
link |
and I don't claim to have mastered it at all,
link |
but I think it's so fascinating about where that will go
link |
with new social networks.
link |
See, what I'm also fascinated by is metrics
link |
that are different than engagement.
link |
So the other thing from an alien perspective,
link |
what social networks are doing is they,
link |
they in the short term,
link |
bring out different aspects of each human being.
link |
So first, let me say that an algorithm or a social network
link |
for each individual can bring out the best of that person
link |
or the worst of that person,
link |
or there's a bunch of different parts to us,
link |
parts we're proud of that we are,
link |
parts we're not so proud of.
link |
When we look at the big picture of our lives,
link |
when we look back 30 days from now,
link |
am I proud that I said those things or not?
link |
Am I proud that I felt those things?
link |
Am I proud that I experienced or read those things
link |
or thought about those things?
link |
Just in that kind of self reflected kind of way.
link |
And so coupled with that,
link |
I wonder if it's possible to have different metrics
link |
that are not just about engagement,
link |
but are about long term happiness,
link |
growth of a human being,
link |
where they look back and say,
link |
I am a better human being
link |
for having spent 100 hours on that app.
link |
And that feels like it's actually strongly correlated
link |
with engagement in the long term.
link |
In the short term, it may not be,
link |
but in long term, it's like the same kind of thing
link |
where you really fall in love with the product.
link |
You fall in love with an iPhone,
link |
you fall in love with a car.
link |
That's what makes you fall in love
link |
is like really being proud
link |
and just in a self reflected way,
link |
understanding that you're a better human being
link |
for having used the thing.
link |
And that's what great relationships are made from.
link |
It's not just like you're hot
link |
and we like being together or something like that.
link |
It's more like I'm a better human being
link |
because I'm with you.
link |
And that feels like a metric
link |
that could be optimized for by the algorithms.
link |
But anytime I kind of talk about this with anybody,
link |
they seem to say, yeah, okay,
link |
that's going to get out competed immediately
link |
by the engagement if it's ad driven especially.
link |
I just don't think so.
link |
I don't, I mean, a lot of it's just implementation.
link |
I'll say a couple of things.
link |
One is to pull back the curtain on daily meetings
link |
inside of these large social media companies.
link |
A lot of what management,
link |
or at least the people that are tweaking these algorithms
link |
spend their time on are trade offs.
link |
And there's these things called value functions,
link |
which are like, okay,
link |
we can predict the probability that you'll click
link |
on this thing or the probability that you'll share it,
link |
or the probability that you will leave a comment on it
link |
or the probability you'll dwell on it.
link |
Individual actions, right?
link |
And you've got this neural network
link |
that basically has a bunch of heads at the end
link |
and all of them are between zero and one and great.
link |
They all have values, right?
link |
Or they all have probabilities.
link |
And then in these meetings, what they will do is say,
link |
well, how much do we value a comment versus a click
link |
versus a share versus a,
link |
and then maybe even some downstream thing, right?
link |
That has nothing to do with the item there,
link |
but like driving follows or something.
link |
And what typically happens is they will say,
link |
well, what are our goals for this quarter at the company?
link |
Oh, we wanna drive sharing up, okay.
link |
Well, let's turn down these metrics
link |
and turn up these metrics.
link |
And they blend them right into a single scalar
link |
which they're trying to optimize.
link |
That is really hard because invariably
link |
you think you're solving for, I don't know,
link |
something called meaningful interactions, right?
link |
This was the big Facebook pivot.
link |
And I don't actually have any internal knowledge.
link |
Like I wasn't in those meetings,
link |
but at least from what we've seen over the last month or so,
link |
it seems by actually trying to optimize
link |
for meaningful interactions,
link |
it had all these side effects of optimizing
link |
for these other things.
link |
And I don't claim to fully understand them,
link |
but what I will say is that trade offs abound.
link |
And as much as you'd like to solve for one thing,
link |
if you have a network of over a billion people,
link |
you're gonna have unintended consequences either way.
link |
And it gets really hard.
link |
So what you're describing is effectively a value model
link |
that says like, can we capture,
link |
this is the thing that I spent a lot of time thinking about,
link |
like, can you capture utility
link |
in a way that like actually measures someone's happiness
link |
that isn't just a, what do they call it?
link |
A surrogate problem where you say, well,
link |
kind of think like the more you use the product,
link |
the happier you are.
link |
That was always the argument at Facebook, by the way.
link |
It was like, well, people use it more,
link |
so they must be more happy.
link |
Turns out there are like a lot of things you use more
link |
that make you less happy in the world.
link |
Not talking about Facebook,
link |
just let's think about whether it's gambling or whatever,
link |
like that you can do more of,
link |
but doesn't necessarily make you happier.
link |
So the idea that time equals happiness,
link |
obviously you can't map utility and time together easily.
link |
There are a lot of edge cases.
link |
So when you look around the world and you say,
link |
well, what are all the ways we can model utility?
link |
That is like one of the,
link |
please, if you know someone smart doing this,
link |
introduce me because I'm fascinated by it.
link |
And it seems really tough.
link |
But the idea that reinforcement learning,
link |
like everyone interesting I know in machine learning,
link |
like I was really interested in recommender systems
link |
and supervised learning.
link |
And the more I dug into it, I was like,
link |
oh, literally everyone smart
link |
is working on reinforcement learning.
link |
Like literally everyone.
link |
You just made people at OpenAI and DeepMind very happy, yes.
link |
But I mean, but what's interesting is like,
link |
it's one thing to train a game and like,
link |
I mean that paper where they just took Atari
link |
and they used a ConvNet to basically just like
link |
train simple actions, mind blowing, right?
link |
Absolutely mind blowing, but it's a game, great.
link |
So now what if you're constructing a feed for a person,
link |
Like how can you construct that feed in such a way
link |
that optimizes for a diversity of experience,
link |
a longterm happiness, right?
link |
But that reward function,
link |
it turns out in reinforcement learning again,
link |
as I've learned, like reward design is really hard.
link |
And I don't know, like how do you design a scalar reward
link |
for someone's happiness over time?
link |
I mean, do you have to measure dopamine levels?
link |
Like, do you have to?
link |
Well, you have to have a lot more signals
link |
from the human being.
link |
Currently it feels like there's not enough signals
link |
coming from the human being users of this algorithm.
link |
So for reinforcement learning to work well,
link |
you need to have a lot more data.
link |
Needs to have a lot of data.
link |
And that actually is a challenge for anyone
link |
who wants to start something,
link |
which is you don't have a lot of data.
link |
So how do you compete?
link |
But I do think back to your original point,
link |
rethinking the algorithm, rethinking reward functions,
link |
rethinking utility, that's fascinating.
link |
And I think that's an open opportunity
link |
for a company that figures it out.
link |
I have to ask about April, 2012,
link |
when Instagram, along with its massive employee base
link |
of 13 people was sold to Facebook for $1 billion.
link |
What was the process like on a business level,
link |
engineering level, human level?
link |
What was that process of selling to Facebook like?
link |
What did it feel like?
link |
So I want to provide some context,
link |
which is I worked in corporate development at Google,
link |
which not a lot of people know,
link |
but corporate development is effectively the group
link |
that buys companies, right?
link |
You sit there and you acquire companies.
link |
And I had sat through so many of these meetings
link |
with entrepreneurs.
link |
We actually, fun fact, we never acquired a single company
link |
when I worked in corporate development.
link |
So I can't claim that I had like a lot of experience,
link |
but I had enough experience to understand,
link |
okay, like what prices are people getting
link |
and what's the process?
link |
And as we started to grow,
link |
we were trying to keep this thing running
link |
and we were exhausted and we were 13 people.
link |
And I mean, we were trying to think back,
link |
it was probably 27, 37 now,
link |
so young on a relative basis, right?
link |
And we're trying to keep the thing running.
link |
And then we go out to raise money
link |
and we're kind of like the hot startup at the time.
link |
And I remember going into a specific VC and saying,
link |
our terms we're looking for are,
link |
we're looking for a $500 million valuation.
link |
And I've never seen so many jaws drop all in unison, right?
link |
And I was like, thanked and walked out the door
link |
very kindly after.
link |
And then I got a call the next day
link |
from someone who was connected to them.
link |
And they said, we just wanna let you know
link |
that like it was pretty offensive
link |
that you asked for a $500 million valuation.
link |
And I can't tell if that was like just negotiating or what,
link |
but it's true, like no one offered us more, right?
link |
So can you clarify the number again?
link |
You said how many million?
link |
500 million, yeah, half a billion.
link |
So in my mind, I'm anchored like, okay,
link |
well, literally no one's biting at 500 million.
link |
And eventually we would get Sequoia and Greylock
link |
and others together at 500 million, basically, post.
link |
It was 450 pre, I think we raised $50 million.
link |
But just like no one was used to seeing
link |
a $500 million companies then.
link |
Like, I don't know if it was because we were just coming
link |
out of the hangover of 2008
link |
and things were still on recovery mode.
link |
But then along comes Facebook.
link |
And after some negotiation, we've two X to the number
link |
from a half a billion to a billion.
link |
Yeah, it seems pretty good.
link |
And I think Mark and I really saw eye to eye
link |
that this thing could be big.
link |
We thought we could...
link |
Their resources would help us scale it.
link |
And in a lot of ways it de risks.
link |
I mean, it de risks a lot of the employees lives
link |
for the rest of their lives,
link |
including me, including Mike, right?
link |
I think I might've had like 10 grand
link |
in my bank account at the time, right?
link |
Like we're working hard, we had nothing.
link |
So on a relative basis, it seems very high.
link |
And then I think the last company to exit
link |
for anywhere close to a billion was YouTube
link |
that I could think of.
link |
And thus began the giant long bull run of 2012
link |
to all the way to where we are now,
link |
where I saw some stat yesterday
link |
about like how many unicorns exist and it's absurd.
link |
But then again, never underestimate technology
link |
and like the value it can provide.
link |
And man, costs have dropped and man scale has increased.
link |
And you can make businesses make a lot of money now.
link |
But on a fundamental level, I don't know,
link |
like how do you describe the decision
link |
to sell a company with 13 people for a billion dollars?
link |
So first of all, like how did it take a lot of guts
link |
to sit at a table and say 500 million
link |
or 1 billion with Mark Zuckerberg?
link |
It seems like a very large number with 13.
link |
Like, especially...
link |
It doesn't seem, it is.
link |
They're all large numbers.
link |
Especially like you said before the unicorn parade.
link |
I like that, I'm gonna use that.
link |
The unicorn parade?
link |
You were at the head of the unicorn parade.
link |
It's the, yeah, it's a massive unicorn parade.
link |
Okay, so no, I mean, we knew we were worth
link |
quote unquote a lot, but we didn't,
link |
I mean, there was no market for Instagram.
link |
I mean, it's not, you couldn't mark to market this thing
link |
in the public markets.
link |
You didn't quite understand what it would be worth
link |
or was worth at the time.
link |
So in a market, an illiquid market
link |
where you have one buyer and one seller
link |
and you're going back and forth,
link |
and well, I guess there were like VC firms
link |
who were willing to invest at a certain valuation.
link |
So I don't know, you just go with your gut.
link |
And at the end of the day, I would say
link |
the hardest part of it was not realizing,
link |
like when we sold, it was tough
link |
because like literally everywhere I go,
link |
restaurants, whatever, like for a good six months after,
link |
there was a lot of attention on the deal,
link |
a lot of attention on the product,
link |
a lot of attention, it was kind of miserable, right?
link |
And you're like, wait, like I made a lot of money,
link |
but like, why is this not great?
link |
And it's because it turns out,
link |
I don't know, like I don't really keep in touch with Mark,
link |
but I've got to assume his job right now
link |
is not exactly the most happy job in the world.
link |
It's really tough when you're on top
link |
and it's really tough when you're in the limelight.
link |
So the decision itself was like, oh, cool, this is great.
link |
How lucky are we, right?
link |
So, okay, there's a million question I want to ask.
link |
First of all, why is it hard to be on top?
link |
Why did you not feel good?
link |
Like, can you dig into that?
link |
It always, I've heard like Olympic athletes say
link |
after they win gold, they get depressed.
link |
Is it something like that where it feels like
link |
it was kind of like a thing you were working towards?
link |
Some loose definition of success.
link |
And this sure as heck feels like
link |
at least according to other startups,
link |
this is what success looks like.
link |
And now why don't I feel any better?
link |
I'm still human, I still have all the same problems.
link |
Is that the nature?
link |
Or is it just like negative attention of some kind?
link |
I think it's all of the above.
link |
But to be clear, there was a lot of happiness
link |
in terms of like, oh my God, this is great.
link |
Like we won the Super Bowl of startups, right?
link |
Anyone who can get to a liquidity event
link |
of anything meaningful feels like,
link |
wow, this is what we started out to do.
link |
Of course we want to create great things that people love,
link |
but like we won in a big way.
link |
But yeah, there's this big like,
link |
oh, if we won, what's next?
link |
So they call it the we have arrived syndrome,
link |
which I need to go back and look where I can quote that from.
link |
But I remember reading about it at the time.
link |
I was like, oh yeah, that's that.
link |
And I remember we had a product manager leave very early on
link |
when we got to Facebook.
link |
And he said to me,
link |
I just don't believe I can learn anything
link |
at this company anymore.
link |
It's like, it's hit its apex.
link |
I just don't have anything else to learn.
link |
So from 2012 all the way to the day I left in 2018,
link |
like the amount I learned and the humility
link |
with which I realized, oh, we thought we won.
link |
Billion dollars is cool,
link |
but like there are a hundred billion dollar companies.
link |
And by the way, on top of that, we had no revenue.
link |
We had, I mean, we had a cool product,
link |
but we didn't scale it yet.
link |
And there's so much to learn.
link |
And then competitors and how fun was it to fight Snapchat?
link |
Like it was, it's like Yankees Red Sox.
link |
Like that's what you live for.
link |
You know, you win some, you lose some,
link |
but the amount you can learn through that process,
link |
what I've realized in life is that there is no,
link |
and there's always someone who has more,
link |
there's always more challenge, just at different scales.
link |
And it sounds like a little Buddhist,
link |
but everything is super challenging,
link |
whether you're like a small business
link |
or an enormous business.
link |
I say like choose the game you like to play, right?
link |
You've got to imagine that
link |
if you're an amazing basketball player,
link |
you enjoy to some extent practicing basketball.
link |
It's gotta be something you love.
link |
It's gonna be hard.
link |
You're gonna have injuries, right?
link |
But you gotta love it.
link |
And the same thing with Instagram,
link |
which is we might've sold, but it was like, great.
link |
There's one Super Bowl title.
link |
What else can we do?
link |
Now I imagine you didn't ask this, but okay, so I left.
link |
There's a little bit of like, what do you do next, right?
link |
Like, how do you top that thing?
link |
It's the wrong question.
link |
The question is like, when you wake up every day,
link |
what is the hardest, most interesting thing
link |
you can go work on?
link |
Because like at the end of the day,
link |
we all turn into dirt, it doesn't matter, right?
link |
But what does matter is like,
link |
can we really enjoy this life?
link |
Not in a hedonistic way, because that's those,
link |
it's like the reinforcement learning,
link |
learning like short term versus long term objectives.
link |
Can you wake up every day and truly enjoy what you're doing
link |
knowing that it's gonna be painful?
link |
Knowing that like, no matter what you choose,
link |
it's gonna be painful.
link |
Whether you sit on a beach
link |
or whether you manage a thousand people or 10,000,
link |
it's gonna be painful.
link |
So choose something that's fun to have pain.
link |
But yes, there was a lot of, we have arrived
link |
and it's a maturation process you just have to go through.
link |
So no matter how much success there is,
link |
how much money you make,
link |
you have to wake up the next day and choose the hard life,
link |
whatever that means next, that's fun.
link |
The fun slash hard life, hard life that's fun.
link |
I guess what I'm trying to say is slightly different,
link |
which is just that no one realizes
link |
everything's gonna be hard.
link |
Even chilling out is hard.
link |
And then you just start worrying about stupid stuff.
link |
Like, I don't know, like did so and so
link |
forget to paint the house today
link |
or like did the gardener come or whatever?
link |
Like, or, oh, I'm so angry
link |
and my shipment of wine didn't show up
link |
and I'm sitting here on the beach without my wine.
link |
I don't know, I'm making shit up now, but like.
link |
It turns out that even chilling, AKA meditation,
link |
Yeah, and at least meditation is like productive chilling
link |
where you're like actually training yourself
link |
to calm down and be, but backing up for a moment,
link |
everything's hard.
link |
You might as well be like playing the game you love to play.
link |
I just like playing and winning and I'm on the,
link |
I'm still on the, I think the first half of life,
link |
knock on wood, and I've got a lot of years
link |
and what am I gonna do, sit around?
link |
And the other way of looking at this, by the way,
link |
imagine you made one movie and it was great.
link |
Would you just like stop making movies?
link |
No, generally you're like, wow,
link |
I really like making movies, let's make another one.
link |
A lot of times, by the way,
link |
the second one or the third one, not that great,
link |
but the fourth one, awesome.
link |
And no one forgets the second,
link |
or everyone forgets the second and the third one.
link |
So there's just this constant process of like,
link |
can I produce and is that fun?
link |
What else can I learn?
link |
So this machine learning stuff for me
link |
has been this awesome new chapter of being like,
link |
man, that's something I didn't understand at all.
link |
And now I feel like I'm one 10th of the way there.
link |
And that feels like a big mountain to climb.
link |
So I distracted us from the original question.
link |
No, and we'll return to the machine learning
link |
cause I'd love to explore your interest there.
link |
But I mean, speaking of sort of challenges and hard things,
link |
is there a possible world
link |
where sitting in a room with Mark Zuckerberg
link |
with a $1 billion deal, you turn it down?
link |
What does that world look like?
link |
Why would you turn it down?
link |
Why did you take it?
link |
What was the calculation that you were making?
link |
Thus enters the world of counterfactuals
link |
and not really knowing.
link |
And if only we could run that experiment.
link |
Well, the universe exists,
link |
it's just running in parallel to our own.
link |
Yeah, it's so fascinating, right?
link |
I mean, we're talking a lot about money,
link |
but the real question was,
link |
I'm not sure you'll believe me when I say this,
link |
but could we strap our little company
link |
onto the side of a rocket ship
link |
and like get out to a lot of people really, really quickly
link |
with the support, with the talent of a place like Facebook?
link |
I mean, people often ask me what I would do differently
link |
at Instagram today.
link |
And I say, well, I'd probably hire more carefully
link |
because we showed up just like before I knew it,
link |
we had like a hundred people on the team
link |
and 200, then 300.
link |
I don't know where all these people were coming from.
link |
I never had to recruit them.
link |
I never had to screen them.
link |
They were just like internal transfers, right?
link |
So it's like relying on the Facebook hiring machine,
link |
which is quite sort of, I mean, it's an elaborate machine.
link |
It's great, by the way.
link |
They have really talented people there.
link |
But my point is the choice was like, take this thing,
link |
put it on the side of a rocket ship
link |
that you know is growing very quickly.
link |
Like I had seen what had happened
link |
when Ev sold Blogger to Google and then Google went public.
link |
Remember we sold before Facebook went public.
link |
There was a moment at which the stock price was $17,
link |
by the way, Facebook stock price was $17.
link |
I remember thinking, what the fuck did I just do, right?
link |
Now at 320 ish, I don't know where we are today.
link |
But like, okay, like the best thing by the way
link |
is like when the stock is down, everyone calls you a dope.
link |
And then when it's up, they also call you a dope,
link |
but just for a different reason, right?
link |
Like you can't win.
link |
Less than in there somewhere.
link |
So, but you know, the choice is to strap yourself
link |
to a rocket ship or to build your own.
link |
You know, Mr. Elon built his own
link |
literally with a rocket ship.
link |
That's a difficult choice because there's a world.
link |
Actually, I would say something different,
link |
which is Elon and others decided to sell PayPal
link |
for not that much.
link |
I mean, how much was it about a billion dollars?
link |
Something like that, yeah.
link |
I mean, it was early and,
link |
but it's worth a lot more now.
link |
To then build a new rocket ship.
link |
So this is the cool part, right?
link |
If you are an entrepreneur
link |
and you own a controlling stake in the company,
link |
not only is it really hard to do something else
link |
with your life because all of the, you know,
link |
value is tied up in you as a personality
link |
attached to this company, right?
link |
But if you sell it and you're getting yourself enough capital
link |
and you like have enough energy,
link |
you can do another thing or 10 other things
link |
or in Elon's case, like a bunch of other things.
link |
I don't know, like I lost count at this point.
link |
And it might've seemed silly at the time.
link |
And sure, like if you look back,
link |
man, PayPal is worth a lot now, right?
link |
Like, do you think Elon like cares about like,
link |
are we gonna buy Pinterest or not?
link |
Like, I just, he is,
link |
he created a massive capital
link |
that allowed him to do what he wants to do.
link |
And that's awesome.
link |
That's more freeing than anything
link |
because when you are an entrepreneur attached to a company,
link |
you gotta stay at that company for a really long time.
link |
It's really hard to remove yourself.
link |
But I'm not sure how much he loved
link |
PayPal versus SpaceX and Tesla.
link |
I have a sense that you love Instagram.
link |
Yeah, I loved enough to like work
link |
for six years beyond the deal.
link |
Which is rare, which is very rare.
link |
But can I tell you why?
link |
There are not a lot of companies that you can be part of
link |
where the Pope's like,
link |
I would like to sign up for your product.
link |
Like I'm not a religious person at all.
link |
You go to the Vatican and you're like walking
link |
among giant columns and you're hearing the music
link |
and everything and like the Pope walks in
link |
and he wants to press the signup button on your product.
link |
Like it's a moment in life, okay?
link |
No matter what your persuasion, okay?
link |
The number of doors and experiences that that opened up
link |
was, it was incredible.
link |
I mean, the people I got to meet, the places I got to go,
link |
I assume maybe like a payments company
link |
is slightly different, right?
link |
But that's why, like it was so fun.
link |
And plus I truly believed we were building
link |
such a great product and I loved, loved the game.
link |
It wasn't about money.
link |
It was about the game.
link |
Do you think you had the guts to say no?
link |
Is that, so here's, I often think about this,
link |
like how hard is it for an entrepreneur to say no?
link |
Because the peer pressure.
link |
So every, like basically the sea of entrepreneurs
link |
in Silicon Valley are gonna tell you,
link |
I mean, this is their dream.
link |
The thing you were sitting before is a dream.
link |
To walk away from that is really,
link |
it seems like nearly impossible.
link |
Because Instagram could in 10 years be,
link |
you could talk about Google.
link |
You could be making self driving cars
link |
and building rockets that go to Mars
link |
and compete with SpaceX.
link |
And so that's an interesting decision to say,
link |
am I willing to risk it?
link |
And the reason I also say it's an interesting decision
link |
because it feels like per our previous discussion,
link |
if you're launching a social network company,
link |
there's going to be that meeting, whatever that number is.
link |
If you're successful, if you're on this rocket ship
link |
of success, there's going to be a meeting
link |
with one of the social media, social network companies
link |
that wanna buy you, whether it's Facebook or Twitter,
link |
but it could also very well be Google
link |
who seems to have like a graveyard
link |
of failed social networks.
link |
And it's, I mean, I don't know.
link |
I think about that, how difficult it is
link |
for an entrepreneur to make that decision.
link |
How many have successfully made that decision?
link |
I guess, this is a big question.
link |
It's sad to me, to be honest,
link |
that too many make that decision,
link |
perhaps for the wrong reason.
link |
Sorry, when you say make the decision,
link |
you mean to the affirmative.
link |
To the affirmative, yeah.
link |
There are also companies that don't sell, right?
link |
And take the PATH and say, we're gonna be independent.
link |
And then you've never heard of them again.
link |
Like I remember PATH, right?
link |
Was one of our competitors early on.
link |
There was a big moment when they had,
link |
I can't remember what it was,
link |
like $110 million offer from Google or something.
link |
It might've been larger, I don't know.
link |
And I remember there was like this big TechCrunch article
link |
that was like, they turned it down after talking deeply
link |
about their values and everything.
link |
And I don't know the inner workings of Foursquare,
link |
but I'm certain there were many conversations over time
link |
where there were companies that wanted Foursquare as well.
link |
Recently, I mean, what other companies?
link |
There's Clubhouse, right?
link |
Like, I don't know.
link |
Maybe people were really interested in them too.
link |
Like there are plenty of moments where people say no
link |
and we just forget that those things happen.
link |
We only focus on the ones where like they said yes
link |
and like, wow, like what if they had stayed independent?
link |
I used to think a lot about this and now I just don't
link |
because I'm like, whatever.
link |
Things have gone pretty well.
link |
I'm ready for the next game.
link |
I mean, think about an athlete where, I don't know,
link |
maybe they do something wrong in the World Series
link |
If you let it haunt you for the rest of your career,
link |
like why not just be like, I don't know, it was a game.
link |
Next game, next shot, right?
link |
And if you just move to that world,
link |
like at least I have a next shot, right?
link |
No, that's beautiful, but I mean, just insights
link |
and it's funny you brought up Clubhouse.
link |
It seems like Clubhouse is on the downward path
link |
and it's very possible to see a billion plus dollar deal
link |
at some stage, maybe like a year ago or half a year ago
link |
from Facebook, from Google.
link |
I think Facebook was flirting with that idea too.
link |
And I think a lot of companies probably were.
link |
I wish it was more public.
link |
There's not like a bad public story
link |
about them making the decision to walk away.
link |
We just don't hear about it.
link |
And then we get to see the results of that success
link |
or the failure, more often failure.
link |
So a couple of things, one is,
link |
I would not assume Clubhouse is down for the count at all.
link |
They're young, they have plenty of money,
link |
they're run by really smart people.
link |
I'd give them like a very fighting chance to figure it out.
link |
There are a lot of times when people call Twitter
link |
down for the count and they figure it out
link |
and they seem to be doing well, right?
link |
So just backing up like,
link |
and not knowing anything about their internals,
link |
like there's a strong chance they will figure it out
link |
and that people are just down
link |
because they like being down about companies.
link |
They like assuming that they're gonna fail.
link |
So who knows, right?
link |
But let's take the ones in the past
link |
where like we know how it played out.
link |
There are plenty of examples
link |
where people have turned down big offers
link |
and then you've just never heard from them again,
link |
but we never focus on the companies
link |
because you just forget that those were big.
link |
But inside your psyche,
link |
I think it's easy for someone with enough money
link |
to say money doesn't matter, which I think is like,
link |
Of course, money matters to people, but at the moment,
link |
you just can't even grasp like the number of zeros
link |
that you're talking about.
link |
It just doesn't make sense, right?
link |
So to think rationally in that moment
link |
is not something many people are equipped to do,
link |
especially not people
link |
where I think we had founded the company a year earlier,
link |
maybe two years earlier, like a year and a half,
link |
we were 13 people, but I will say,
link |
I still don't know if it was the right decision
link |
because I don't have that counterfactual.
link |
I don't know that other world.
link |
I'm just thankful that by and large,
link |
most people love Instagram, still do.
link |
By and large, people are very happy
link |
with like the time we had there
link |
and I'm proud of what we built.
link |
So like, I'm cool.
link |
Like now it's next shot, right?
link |
Well, if we could just linger on this Yankees versus Red Sox,
link |
the fun of it, the competition over,
link |
I would say over the space of features.
link |
So there are a bunch of features,
link |
like there's photos, there's one minute videos on Instagram,
link |
there's IGTV, there's stories, there's reels, there's live.
link |
So that sounds like it's like a long list
link |
of too much stuff, but it's not
link |
because it feels like they're close together,
link |
but they're somehow, like what we're saying,
link |
fundamentally distinct,
link |
like each of the things I mentioned.
link |
Maybe can you describe the philosophies,
link |
the design philosophies behind some of these,
link |
how you were thinking about it
link |
during the historic war between Snapchat and Instagram
link |
or just in general,
link |
like this space of features that was discovered.
link |
There's this great book by Clay Christensen
link |
called, Competing Against Luck.
link |
It's like a terrible title,
link |
but within it, there's effectively an expression
link |
of this thing called jobs to be done theory.
link |
And it's unclear if like he came up with it
link |
or some of his colleagues,
link |
but there are a bunch of places you can find
link |
with people claiming to have come up
link |
with this jobs to be done theory.
link |
But the idea is if you like zoom out
link |
and you look at your product,
link |
you ask yourself, why are people hiring your product?
link |
Like imagine every product in your life
link |
is effectively an employee, you know,
link |
you're CEO of your life
link |
and you hire products to be employees effectively.
link |
They all have roles and jobs, right?
link |
Why are you hiring a product?
link |
Why do you want that product
link |
to perform something in your life?
link |
And like, what are the hidden reasons
link |
why you're in love with this product?
link |
Instagram was about sharing your life
link |
with others visually, period, right?
link |
Why? Because you feel connected with them.
link |
You get to show off.
link |
You get to feel good and cared about, right?
link |
With likes and it turns out that that will,
link |
I think forever define Instagram
link |
and any product that serves that job
link |
is going to do very well, okay?
link |
Stories let's take as an example
link |
is very much serving that job.
link |
In fact, it serves it better than the original product
link |
because when you're large and have an enormous audience,
link |
you're worried about people seeing your stuff
link |
or you're worried about being permanent
link |
so that a college admissions person
link |
is going to see your photo of you doing something.
link |
And so it turns out that that is a more efficient way
link |
of performing that job than the original product was.
link |
The original product still has its value,
link |
but at scale, these two things together
link |
work really, really well.
link |
Now, I will claim that other parts of the product
link |
over time didn't perform that job as well.
link |
I think IGTV probably didn't, right?
link |
Shopping is like completely unrelated
link |
to what I just described, but it might work.
link |
I don't know, right?
link |
Products I think, products that succeed
link |
are products that all share this parent node
link |
of like this job to be done that is in common.
link |
And then they're just like different ways of doing it, right?
link |
Apple, I think does a great job with this, right?
link |
It's like managing your digital life
link |
and all the products just work together.
link |
They sync, they're like, it's beautiful, right?
link |
Even if they require like silly specific cords to work,
link |
but they're all part of a system.
link |
It's when you leave that system
link |
and you start doing something weird
link |
that people start scratching their head
link |
and I think you are less successful.
link |
So I think one of the challenges
link |
Facebook has had throughout its life
link |
is that it has never fully, I think,
link |
appreciated the job to be done of the main product.
link |
And what it's done is said,
link |
ooh, there's a shiny object over there.
link |
That startup's getting some traction.
link |
Let's go copy that thing.
link |
And then they're confused why it doesn't work.
link |
Like why doesn't it work?
link |
It's because the people who show up for this
link |
don't want that, it's different.
link |
What's the purpose of Facebook?
link |
So I remember I was a very early Facebook user.
link |
The reason I was personally excited about Facebook
link |
is because you can, first of all, use your real name.
link |
Like I can exist in this world.
link |
It can be like formally exist.
link |
I like anonymity for certain things, Reddit and so on,
link |
but I want it to also exist not anonymously
link |
so that I can connect with other friends of mine
link |
nonanonymously and there's a reliable way to know
link |
that I'm real and they're real and that we're connecting.
link |
And it's kind of like, I liked it for the reasons
link |
that people like LinkedIn, I guess.
link |
But like without the form, like not everybody is dressed up
link |
and being super polite, like more like with friends.
link |
But then it became something much bigger than that,
link |
I suppose, there's a feed.
link |
It became this, I mean, it became a place
link |
to discover content, to share content
link |
that's not just about connecting directly with friends.
link |
I mean, it became something else.
link |
I don't even know what it is really.
link |
So you said Instagram is a place
link |
where you visually share your life.
link |
Well, let's go back to the founding of Facebook
link |
and why it worked really well initially at Harvard
link |
and then Dartmouth and Stanford and I can't remember,
link |
probably MIT, there were like a handful of schools
link |
in that first tranche, right?
link |
It worked because there are communities
link |
that exist in the world that want to transact.
link |
And when I say transact, I don't mean commercially,
link |
I just mean they want to share, they want to coordinate,
link |
they want to communicate, they want a space for themselves.
link |
And Facebook at its best, I think is that.
link |
And if actually you look at the most popular products
link |
that Facebook has built over time,
link |
if you look at things like groups, marketplace,
link |
groups is enormous.
link |
And groups is effectively like everyone can found
link |
their own little Stanford or Dartmouth or MIT, right?
link |
And find each other and share and communicate
link |
about something that matters deeply to them.
link |
That is the core of what Facebook was built around.
link |
And I think today is where it stands most strongly.
link |
Yeah, it's brilliant.
link |
The groups, I wish groups were done better.
link |
It feels like it's not a first class citizen.
link |
I know I may be saying something without much knowledge,
link |
but it feels like it's kind of bolted on
link |
while being used a lot.
link |
It feels like there needs to be a little bit more structure
link |
in terms of discovery, in terms of like.
link |
I mean, look at Reddit.
link |
Like Reddit is basically groups of public and open
link |
and a little bit crazy, right?
link |
But there's clear product market fit
link |
for that specific use case.
link |
And it doesn't have to be a college, it can be anything.
link |
It can be a small group, a big group,
link |
it can be group messaging.
link |
Facebook shines, I think, when it leans into that.
link |
I think when there are other companies
link |
that just seem exciting and now all of a sudden
link |
the product shifts in some fundamental way
link |
to go try to compete with that other thing,
link |
that's when I think consumers get confused.
link |
Even if you can be successful,
link |
like even if you can compete with that other company,
link |
even if you can figure out how to bolt it on,
link |
eventually you come back and you look at the app
link |
and you're like, I just don't know why I opened this app.
link |
Like why, like too many things going on.
link |
And that was always a worry.
link |
I mean, you listed all the things at Instagram
link |
and it almost gave me a heart attack,
link |
like way too many things.
link |
But I don't know, entrepreneurs get bored.
link |
They want to add things.
link |
They want to like, right?
link |
I don't have a good answer for it,
link |
except for that I think being true to your original use case
link |
and not even original use case,
link |
but sorry, actually not use case, original job.
link |
There are many use cases under that job.
link |
Being true to that and like being really good at it
link |
over time and morphing as needs change,
link |
I think that's how to make a company last forever.
link |
And I mean, honestly, like my main thesis
link |
about why Facebook is in the position it is today
link |
is if they have had a series of product launches
link |
that delighted people over time,
link |
I think they'd be in a totally different world.
link |
So just like imagine for a moment,
link |
and by the way, Apple's entering this,
link |
but like Apple for so long,
link |
just like product after product,
link |
you couldn't wait for it.
link |
You stood in line for it.
link |
You talked about it.
link |
Amazon makes your life so easy.
link |
It's like, wow, I needed this thing
link |
and it showed up at my door two days later.
link |
And like both of these companies, by the way,
link |
Amazon, Apple have issues, right?
link |
There are labor issues,
link |
whether it's here in the US or in China,
link |
there are environmental issues.
link |
But like when's the last time
link |
you heard like a large chorus being like,
link |
these companies better pay
link |
for what they're doing on these things, right?
link |
I think Facebook's main issue today is like,
link |
you need to produce a hit.
link |
If you don't produce hits,
link |
it's really hard to keep consumers on your side.
link |
Then people just start picking on you
link |
for a variety of reasons, whether it's right or wrong.
link |
I'm not even going to place a judgment
link |
right here and right now.
link |
I'm just going to say that it is way better
link |
to be in a world where you are producing hits
link |
and consumers love what you're doing
link |
because then they're on your side.
link |
And I think that's, it's the past 10 years
link |
for Facebook has been fairly hard on this dimension.
link |
So, and by hits, it doesn't necessarily mean financial hits.
link |
It feels like to me, what you're saying
link |
is something that brings joy.
link |
A product that brings joy
link |
to some fraction of the population.
link |
Yeah, I mean, TikTok isn't just literally an algorithm.
link |
In some ways, TikTok's content and algorithm
link |
have more sway now over the American psyche
link |
than Facebook's algorithm, right?
link |
It's visual, it's video.
link |
By the way, it's not defined by who you follow.
link |
It's defined by some magical thing that,
link |
by the way, if someone wanted to tweak
link |
to show you a certain type of content for some reason,
link |
they could, right, but people love it.
link |
So as a CEO, let me ask you a question
link |
because leadership matters.
link |
This is a complicated question.
link |
Why is Mark Zuckerberg distrusted, disliked
link |
and sometimes even hated by many people in public?
link |
Right, that is a complicated question.
link |
Well, the premise, I'm not sure I agree with the premise.
link |
And I can expand that to include
link |
even a more mysterious question for me, Bill Gates.
link |
What is the Bill Gates version of the question?
link |
Do you think people hate Bill Gates?
link |
So takeaway one, it's a checklist.
link |
There is, I think Mark Zuckerberg's distrust
link |
is the primary one, but there's also like a dislike,
link |
maybe hate is too strong a word,
link |
but it's just if you look at like the articles
link |
that are being written and so on, there's a dislike.
link |
And it makes, it's confusing to me
link |
because it's like the public picks certain individuals
link |
and they attach certain kinds of emotions
link |
to those individuals.
link |
Yeah, so someone once just recently said,
link |
there's a strong case that founder led companies
link |
have this problem and that a lot of Mark's issues
link |
come today come from the fact that he is a visible founder
link |
with this story that people have watched in both a movie
link |
and they followed along and he's this boy wonder kid
link |
who became one of the world's richest people
link |
and he's no longer Mark the person,
link |
he's Mark this image of a person
link |
with enormous wealth and power.
link |
And in today's world, we have issues
link |
with enormous wealth and power for a variety of reasons.
link |
One of which is we've been stuck inside
link |
for a year and a half, two years.
link |
One of which is a lot of people were really unhappy
link |
about not the last election, but the last, last election.
link |
And where do you take out that anger?
link |
Who do you blame but the people in charge?
link |
That's one example or one reason why I think
link |
a lot of people express anger or resentment
link |
or unhappiness with Mark.
link |
At the same time, I don't know,
link |
I pointed out to that person, I was like, well,
link |
I don't know, I think a lot of people really like Elon.
link |
Elon arguably, he kept his factory open here
link |
throughout COVID protocols, which arguably
link |
a lot of people would be against.
link |
While saying a bunch of crazy offensive things
link |
on the internet, they still.
link |
And basically gives the middle finger to the SEC
link |
on Twitter and I don't know, I'm like,
link |
well, there's a founder and like people kind of like him.
link |
So I do think that the founder and slash CEO
link |
of a company that's a social network company
link |
is like an extra level of difficulty.
link |
If life is a video game,
link |
you just chose the harder video game.
link |
So, I mean, that's why it's interesting to ask you
link |
because you were the founder and CEO of a social network.
link |
I challenge it because.
link |
Exactly, but you're one of the rare examples.
link |
Even Jack Dorsey's disliked, not to the degree,
link |
but it just seems harder
link |
when you're running a social media company.
link |
I never thought of Jack as just like,
link |
I think generally he's well respected.
link |
I think you're right, but he's not loved.
link |
And I feel like you, I mean, to me,
link |
Twitter is an incredible thing.
link |
Yeah, again, can I just come back to this point,
link |
which seems over simplistic,
link |
but I really do think how a product makes someone feel,
link |
they ascribe that feeling to the founder.
link |
So make people feel good.
link |
So think about it.
link |
Let's just go with this thesis for a second.
link |
Sure, I like it though.
link |
Amazon's pretty utilitarian, right?
link |
It delivers brown boxes to your front door.
link |
Sure, you can have Alexa
link |
and you can have all these things, right?
link |
But in general, it delivers stuff quickly to you
link |
at a reasonable price, right?
link |
I think Jeff Bezos is wonderfully wealthy,
link |
thoughtful, smart guy, right?
link |
But people kind of feel that way about him.
link |
They're like, wow, this is really big.
link |
We're impressed that this is really big.
link |
But he's doing the same space stuff Elon's doing,
link |
but they don't necessarily ascribe
link |
the same sense of wonder, right?
link |
Now let's take Elon.
link |
And again, this is pet theory.
link |
I don't have much proof other than my own intuition.
link |
He is literally about living the future.
link |
Mars, it's about wonder.
link |
It's about going back to that feeling as a kid
link |
when you looked up to the stars and asked,
link |
is there life out there?
link |
People get behind that because it's a sense of hope
link |
and excitement and innovation.
link |
And like, you can say whatever you want,
link |
but we ascribe that emotion to that person.
link |
Now, let's say you're on a social network
link |
and people make you kind of angry
link |
because they disagree with you
link |
or they say something ridiculous
link |
or they're living a FOMO type life where you're like,
link |
wow, I wish I was doing that thing.
link |
I think Instagram, if I were to think back,
link |
by and large, when I was there,
link |
was not about FOMO, was not about this influencer economy,
link |
although it certainly became that way closer to the end.
link |
It was about the sense of wonder and happiness
link |
and beautiful things in the world.
link |
And I don't know, I mean, like,
link |
I don't want to have a blind spot,
link |
but I don't think anyone had a strong opinion
link |
about me one way or the other.
link |
For the longest time, the way people explained to me,
link |
I mean, if you want to go for toxicity,
link |
you go to Facebook or Twitter.
link |
If you want to go to feel good about life,
link |
you go to Instagram to enjoy, celebrate life.
link |
And my experience been talking to people
link |
is they gave me the benefit of the doubt because of that.
link |
But if your experience of the product
link |
is kind of makes you angry, it's where you argue.
link |
I mean, a big part of Jack might be
link |
that he wasn't actually the CEO for a very long time
link |
and only became recently.
link |
So I'm not sure how much of the connection got made.
link |
But in general, I mean, if you hate,
link |
I'm just thinking about other companies
link |
that aren't tech companies.
link |
If you hate like what a company is doing
link |
or it makes you not feel happy.
link |
I don't know, like people are really angry
link |
about Comcast or whatever.
link |
Are they even called Comcast anymore?
link |
It's like Xfinity or something, right?
link |
They had to rebrand.
link |
They became meta, right?
link |
And it's like, but my point is if it makes you angry.
link |
That's beautiful, yeah.
link |
But the thing is, this is me saying this.
link |
I think your thesis is very strong and correct,
link |
has elements of correctness,
link |
but I still personally put some blame on individuals.
link |
I think you said, Elon, looking up,
link |
there's something about Childlike Wander to him,
link |
like to his personality, his character,
link |
something about, I think more so than others
link |
where people can trust them.
link |
And there's, I don't know,
link |
Sondra Prachai is an example of somebody who's like,
link |
there's some kind, it's hard to put into words,
link |
but there's something about the human being
link |
where he's trustworthy.
link |
He's human in a way that connects to us.
link |
And the same with Sajid Nadella.
link |
I mean, some of these folks, something about us
link |
is drawn to them, even when they're flawed.
link |
Even like, so your thesis really holds up for Steve Jobs
link |
because I think people didn't like Steve Jobs,
link |
but he delivered products
link |
and then they fell in love every time.
link |
I guess you could say that the CEO,
link |
the leader is also a product.
link |
And if they keep delivering a product that people like
link |
by being in public and saying things that people like,
link |
that's also a way to make people happy.
link |
But from a social network perspective,
link |
it makes me wonder how difficult it is
link |
to explain to people why certain things happen,
link |
like to explain machine learning,
link |
to explain why certain,
link |
the woke mob effect happens
link |
or the certain kinds of like bullying happens,
link |
which is like, it's human nature combined with algorithm.
link |
And it's very difficult to control for
link |
how the spread of quote unquote misinformation happens.
link |
It's very difficult to control for that.
link |
And so you try to decelerate certain parts
link |
and you create more problems than you solve.
link |
And anything that looks at all like censorship
link |
can create huge amounts of problems as a slippery slope.
link |
And then you have to inject humans
link |
to oversee the machine learning algorithms.
link |
And anytime you inject humans into the system,
link |
it's gonna create a huge number of problems.
link |
And I feel like it's up to the leader
link |
to communicate that effectively, to be transparent.
link |
First of all, design products
link |
that don't have those problems.
link |
And second of all, when they have those problems,
link |
to be able to communicate with them.
link |
I guess that's all going to,
link |
when you run a social network company, your job is hard.
link |
Yeah, I will say the one element that you haven't named
link |
that I think you're getting at is just bedside manner,
link |
which Steve Jobs, I never worked for him.
link |
I never met him in person.
link |
Had an uncanny ability in public to have bedside manner.
link |
I mean, some of the best clips of Steve Jobs
link |
from like, I would say maybe the 80s
link |
when he's on the stage and getting questions
link |
from the audience about life.
link |
And he'll take this question that is like,
link |
how are you gonna compete with blah?
link |
And it's super boring.
link |
And I don't even know the name of the company.
link |
And his answer is as if you just asked your grandfather,
link |
the meaning of life.
link |
And you sit there and you're just like, what?
link |
And there's that bedside manner.
link |
And if you lack that, or if that's just not intuitive to you,
link |
I think that it can be a lot harder
link |
to gain the trust of people.
link |
And then add on top of that, missteps of companies.
link |
I don't know if you have any friends from the past
link |
where maybe they crossed you once,
link |
or maybe you get back together and you're friends again,
link |
but you just never really forget that thing.
link |
It's human nature not to forget.
link |
I'm Russian, you crossed me once.
link |
We solved the problem.
link |
So my point is, humans don't forget.
link |
And if there are times in the past
link |
where they feel like they don't trust the company
link |
or the company hasn't had their back,
link |
that is really hard to earn back,
link |
especially if you don't have that bedside manner.
link |
And again, I'm not attributing this specifically to Marc
link |
because I think a lot of the companies have this issue
link |
where one, you have to be trustworthy as a company
link |
and live by it and live by those actions.
link |
And then two, I think you need to be able
link |
to be really relatable in a way that's very difficult
link |
if you're worth what these people are.
link |
Jack does a pretty good job of this by being a monk.
link |
But I also, Jack issues attention.
link |
He's not out there almost on purpose.
link |
He's just working hard, doing square, right?
link |
I literally shared a desk like this with him at Odeo.
link |
I mean, this normal guy who likes painting,
link |
I remember he would leave early on Wednesdays or something
link |
to go to a painting class.
link |
And he's creative, he's thoughtful.
link |
I mean, money makes people more creative and more thoughtful,
link |
extreme versions of themselves, right?
link |
And this was a long, long time ago.
link |
You mentioned that he asked you
link |
to do some kind of JavaScript thing.
link |
We were working on some JavaScript together.
link |
That's hilarious, like pre Twitter, early Twitter days,
link |
you and Jack Dorsey are in a room together
link |
talking about JavaScript,
link |
solving some kind of menial problem.
link |
Terrible problems, yeah.
link |
I mean, not terrible, just like boring widget problem.
link |
I think it was the Odeo widget
link |
we were working on at the time.
link |
I'm surprised anyone paid me to be in the room as an intern
link |
because I didn't really provide any value.
link |
I'm very thankful to anyone who included me back in the day.
link |
It was very helpful.
link |
So thank you for listening.
link |
I mean, is there Odeo that's a precursor to Twitter?
link |
First of all, did you have any anticipation
link |
that this Jack Dorsey guy could be also
link |
a head of a major social network?
link |
And second, did you learn anything from the guy that,
link |
like, do you think it's a coincidence
link |
that you two were in the room together?
link |
And it's the coincidence meaning like,
link |
why does the world play its game in a certain way
link |
where these two founders of social networks?
link |
It's so weird, right?
link |
Like, I mean, it's also weird that Mark showed up
link |
in our fraternity my sophomore year
link |
and we got to know each other then,
link |
like long before Instagram.
link |
It's a small world,
link |
but let me tell a fun story about Jack.
link |
We're at Odeo and I don't know,
link |
I think Ev was feeling like people
link |
weren't working hard enough or something.
link |
And I can't remember exactly what he,
link |
he created this thing where every Friday,
link |
I don't know if it was every Friday,
link |
I only remember this happening once,
link |
but he had us like a statuette, it's like of Mary.
link |
And in the bottom, it's hollow, right?
link |
And I remember on a Friday,
link |
Ev decided he was going to let everyone vote for
link |
who had worked the hardest that week.
link |
We all voted, closed ballot, right?
link |
We all put it in a bucket and he tallied the votes.
link |
And then whoever got the most votes, as I recall,
link |
got the statuette.
link |
And in the statuette was a thousand bucks,
link |
or I recall there was a thousand bucks in it.
link |
It might've been a hundred bucks,
link |
but let's call it a thousand.
link |
It's more exciting that way.
link |
It felt like a thousand, yeah.
link |
It did to me for sure.
link |
I actually got two votes.
link |
We were a small company, but as the intern,
link |
I got at least two votes.
link |
So everybody knew how many votes they got individually?
link |
And I think it was one of these self accountability things.
link |
Anyway, I remember Jack just getting
link |
like the vast majority of votes from everyone.
link |
And I remember just thinking like,
link |
like I couldn't imagine he would become what he'd become
link |
and do what he would do,
link |
but I had a profound respect that the new guy
link |
who I really liked worked that hard.
link |
And you could see his dedication even then
link |
and that people respected him.
link |
That's the one story that I remember of him
link |
like working with him specifically from that summer.
link |
Can I take a small tangent on that?
link |
There's kind of a pushback in Silicon Valley
link |
a little bit against hard work.
link |
Can you speak to the sort of the thing you admire
link |
to see the new guy working so hard, that thing?
link |
What is the value of that thing in a company?
link |
See, this is like, just to be very frank,
link |
it drives me nuts.
link |
Like I saw this really funny video on TikTok.
link |
It was like, I'm taking a break from my mental health
link |
to work on my career.
link |
I thought that was funny.
link |
Um, so I was like, oh, it is kind of phrased that way,
link |
the opposite often, right?
link |
Okay, so a couple of things.
link |
Uh, I, uh, I have worked so hard
link |
to do the things that I did.
link |
Like Mike and I lost years off of our lives,
link |
staying up late, figuring things out,
link |
the stress that comes with the job.
link |
I have a lot more gray hair now than I did back then.
link |
It requires an enormous amount of work
link |
and most people aren't successful, right?
link |
But even the ones that do don't skate by.
link |
I am okay if people choose not to work hard
link |
because I don't actually think there's anything
link |
in this world that says you have to work hard.
link |
But I do think that great things require a lot of hard work.
link |
So there's no way you can expect to change the world
link |
without working really hard.
link |
By the way, even changing the world,
link |
you know, the folks that I respect the most
link |
have nudged the world in like a slight direction,
link |
slight, very, very slight.
link |
Like even if Elon accomplishes all the things
link |
he wants to accomplish,
link |
we will have nudged the world in a slight direction,
link |
but it requires enormous amount.
link |
There was an interview with him where he was just like,
link |
he was interviewed, I think, at the Tesla factory
link |
and he was like, work is really hard.
link |
This is actually unhealthy.
link |
And I can't recall the exact,
link |
but he was like visibly shaken
link |
about how hard he had been working.
link |
And he was like, this is bad.
link |
And unfortunately, I think to have great outcomes,
link |
you actually do need to work
link |
at like three standard deviations above the mean,
link |
but there's nothing saying that people have to go for that.
link |
See, the thing is, but what I would argue,
link |
this is my personal opinion,
link |
is nobody has to do anything, first of all.
link |
They certainly don't have to work hard.
link |
But I think hard work in a company should be admired.
link |
And you should not feel like,
link |
you shouldn't feel good about yourself for not working hard.
link |
Like, so for example, I don't have to work out.
link |
I don't have to run.
link |
I hate running, but like, I certainly don't feel good
link |
if I don't run because I know for my health,
link |
like there's certain values,
link |
I guess is what I'm trying to get at.
link |
There's certain values that you have in life.
link |
It feels like there's certain values
link |
that companies should have in hard work.
link |
Certain values that companies should have in hard work
link |
is one of the things I think that should be admired.
link |
I often ask this kind of silly question,
link |
just to get a sense of people,
link |
like if I'm hiring and so on.
link |
I just ask if they think it's better
link |
to work hard or work smart.
link |
It was helpful for me to get a sense of people from that.
link |
Because you think like the right.
link |
The answer's both.
link |
The answer's both.
link |
The answer's both.
link |
I usually try not to give them that,
link |
but sometimes I'll say both if that's an option.
link |
But a lot of people kind of,
link |
a surprising number will say work smart.
link |
And there are usually people
link |
who don't know how to work smart.
link |
And they're literally just lazy.
link |
Not just, there's two effects behind that.
link |
One is laziness and the other is ego.
link |
When you're younger and you say it's better to work smart,
link |
it means you think you know what it means
link |
to work smart at this early stage.
link |
To me, people that say work hard or both,
link |
they have the humility to understand like,
link |
I'm going to have to work my ass off
link |
because I'm too dumb to know how to work smart.
link |
And people who are self critical in this way,
link |
in some small amount, you have to have some confidence.
link |
But if you have humility,
link |
that means you're going to actually eventually figure out
link |
what it means to work smart.
link |
And then to actually be successful, you should do both.
link |
So I have a very particular take on this,
link |
which is that no one's forcing you to do anything.
link |
All choices have consequences.
link |
So if you major in, I don't know, theoretical literature,
link |
I don't even know if that's a major.
link |
I'm just making something up.
link |
That's supposed to regular literature, applied literature.
link |
Yeah, think about like theoretical Spanish lit
link |
from the 14th century.
link |
Like just make up your esoteric thing.
link |
And then the number of people I went to Stanford with
link |
who get out in the world and they're like,
link |
wait, what, I can't find a job?
link |
Like no one wants a theoretical,
link |
like there are plenty of counter examples
link |
of people who have majored in esoteric things
link |
and gone on to be very successful.
link |
So I just want to be clear, it's not about the major.
link |
But every choice you make, whether it's to have kids,
link |
like I love my children.
link |
It's so awesome to have two kids.
link |
And it is so hard to work really hard and also have kids.
link |
And there's a reason why certain very successful people
link |
like don't have, or not successful,
link |
but people who run very, very large companies or startups
link |
have chosen not to have kids for a while
link |
or chosen not to like prioritize them.
link |
Everything's a choice.
link |
And like I choose to prioritize my children
link |
because like I want to do that, right?
link |
So everything's a choice.
link |
Now, once you've made that choice,
link |
I think it's important that the contract is clear,
link |
let's imagine you were joining a new startup.
link |
It's important that that startup communicate
link |
that like the expectation is like,
link |
we're all working really, really hard right now.
link |
You don't have to join the startup.
link |
But like, if you do just know like you're,
link |
it's almost as if you join, I don't know,
link |
pick your like sports team.
link |
Like let's go back to the Yankees for a second.
link |
You want to join the Yankees,
link |
but you don't really want to work that hard.
link |
You don't really want to do batting practice
link |
or pitching practice or whatever for your position, right?
link |
That to me is wacko.
link |
And that's actually the world that it feels like we live in
link |
in tech sometimes,
link |
where people both want to work for the Yankees
link |
because it pays a lot,
link |
but like don't actually want to work that hard.
link |
That I don't fully understand.
link |
Because if you sign up for some of these things,
link |
just sign up for it.
link |
But it's okay if you don't want to sign up for it.
link |
There's so many wonderful careers in this world
link |
that don't require 80 hours a week.
link |
But when I read about companies going to like
link |
four day work weeks and stuff,
link |
I just like, I chuckle because I can't get enough done
link |
with a seven day week.
link |
And people will say, oh, you're just not working smart.
link |
And it's like, no, I work pretty smart,
link |
I think in general.
link |
Like I wouldn't have gotten to this point
link |
if I hadn't like some amount of working smart.
link |
And there is balance though.
link |
So I used to be like a pretty big cyclist.
link |
I don't do it much anymore just because of kids
link |
and like prioritizing other things, right?
link |
But one of the most important things to learn
link |
as a cyclist is to take a rest day.
link |
But to me and to cyclists,
link |
like resting is a function of optimizing for the long run.
link |
It's not like a thing that you do for its own merits.
link |
It's actually like, if you don't rest,
link |
your muscles don't recover.
link |
And then you're just not as,
link |
like you're not training as efficiently.
link |
You should probably, the successful people I've known
link |
in terms of athletes, they hate rest days,
link |
but they know they have to do it for the long term.
link |
They think their opposition is getting stronger and stronger
link |
and that's the feeling,
link |
but you know it's the right thing
link |
and usually you need a coach to help you.
link |
So, I mean, I use this thing called training peaks
link |
and it's interesting
link |
because it actually mathematically shows
link |
like where you are on the curve and all this stuff,
link |
but you have to have that rest,
link |
but it's a function of going harder for longer.
link |
Again, it's this reinforcement learning,
link |
like planning the aggregate and the long,
link |
but a lot of people will hide behind laziness
link |
by saying that they're trying to optimize for the long run
link |
and they're not, they're just not working very hard.
link |
But again, you don't have to sign up for it.
link |
It's totally cool.
link |
Like, I don't think less of people
link |
for like not working super hard.
link |
It's just like, don't sign up for things
link |
that require working super hard.
link |
And some of that requires for the leadership
link |
to have the guts, the boldness to communicate effectively
link |
at the very beginning.
link |
I mean, sometimes I think most of the problems arise
link |
in the fact that the leadership is kind of hesitant
link |
to communicate the socially difficult truth
link |
of what it takes to be at this company.
link |
So they kind of say, hey, come with us.
link |
There's, we have snacks, you know, but.
link |
Unlimited vacation and.
link |
You know, Ray at Bridgewater is always fascinating
link |
because, you know, people,
link |
it's been called like a cult on the outside
link |
but what's fascinating is like,
link |
they just don't give on their principles.
link |
They're like, listen, this is what it's like to work here.
link |
We record every meeting.
link |
We're like brutally honest
link |
and that's not going to feel right to everyone.
link |
And if it doesn't feel right to you, totally cool.
link |
Just go work somewhere else.
link |
But if you work here, you are signing up for this.
link |
And that's, that's been fascinating to me
link |
because it's honesty upfront.
link |
It's a system in which you operate.
link |
And if it's not for you,
link |
like no one's forcing you to work there, right?
link |
So I did a conversation with him
link |
and kind of got stuck in a funny moment,
link |
which is at the end I asked him, you know,
link |
to give me honest feedback of how I did on the interview.
link |
He was super nice.
link |
He asked me, he's like, well,
link |
tell me, did you accomplish
link |
what you were hoping to accomplish?
link |
I was like, that's not, that's not.
link |
I'm asking you as an objective observer
link |
of two people talking, how do we do today?
link |
And then he's like, well,
link |
he gave me this politician answer.
link |
Well, I feel like we've accomplished
link |
successful communication of like ideas,
link |
which is I'd love to spread some of the ideas
link |
in that, like in principles and so on.
link |
Back to my original point,
link |
it's really hard to get.
link |
It's really hard to give feedback.
link |
And one of the other things I learned from him
link |
and just people in that world is like,
link |
man, humans really like to pretend
link |
like they've come to,
link |
that they've come to some kind of meeting of the minds.
link |
Like if there's conflict, if you and I have conflict,
link |
it's always better to meet face to face, right?
link |
We're on the phone.
link |
Slack is not great, right?
link |
Email is not great, but face to face.
link |
What's crazy is you and I get together
link |
and we actively try to,
link |
even if we're not actually solving the conflict,
link |
we actively try to paper over the conflict.
link |
Oh yeah, it didn't really bother me that much.
link |
Oh yeah, I'm sure you didn't mean it.
link |
But like, no, in our minds we're still there.
link |
So this is one of the things that as a leader,
link |
you always have to be digging,
link |
especially as you ascend.
link |
Like straight to the conflict.
link |
Yeah, but as you ascend,
link |
no one wants to tell you you're crazy.
link |
No one wants to tell you your idea is bad.
link |
And you can, you're like, oh,
link |
oh, I'm going to be a leader.
link |
And the idea is, well, I'm just going to ask people.
link |
So like you have to look for the markers,
link |
knowing that literally just people
link |
aren't going to tell you along the way and be paranoid.
link |
I mean, you asked about selling the company.
link |
I think one of the biggest differences between me
link |
and a lot of other entrepreneurs is like,
link |
I wasn't completely confident we could do it.
link |
Like we could be alone and actually be great.
link |
And if any entrepreneur is honest with you,
link |
they also feel that way.
link |
But a lot of people are like,
link |
well, I have to be cocky and just say,
link |
I can do this on my own.
link |
We're going to be fine.
link |
We're going to crush everyone.
link |
Some people do say that and then it's not right.
link |
And they, and they fail.
link |
But being honest in that moment with yourself,
link |
with those close to you.
link |
And also you talked about the personality of leaders
link |
and who resonates and who doesn't.
link |
It's rare that I see leaders be vulnerable, rare.
link |
And one thing I tried to do at Instagram,
link |
at least internally was like, say when I screwed up
link |
and like point out how I was wrong about things
link |
and point out where my judgment was off.
link |
Everyone thinks they have to bat a thousand, right?
link |
Like that's crazy.
link |
The best quant hedge funds in the world,
link |
They just take a lot of bets, right?
link |
Renaissance, they might, they might bat 51%, right?
link |
But holy hell, like the question isn't,
link |
are you right every single time
link |
and you have to seem invincible.
link |
The question is how many at bats do you get?
link |
And on average, are you better on average, right?
link |
With enough bets and enough at bats
link |
that your aggregate can be very high.
link |
I mean, Steve Jobs was wrong at a lot of stuff.
link |
The Newton too early, right?
link |
Next, not quite right.
link |
There was even a time where he said like,
link |
no one will ever wanna watch a video on the iPod.
link |
But who cares if you come around
link |
and realize your mistake and fix it.
link |
It becomes just like you said, harder and harder
link |
when your ego grows and the number of people around you
link |
that say positive things towards you grows.
link |
I actually think it's really valuable that,
link |
like let's imagine a counterfactual
link |
where Instagram became worth like $300 billion
link |
or something crazy, right?
link |
I kind of like that my life is relatively normal now.
link |
When I say relatively, you get what I mean.
link |
I'm not making a claim that I live a normal life,
link |
but like I certainly don't live in a world
link |
where there are like 15 Sherpas following me,
link |
like fetching me water, like that's not how it works.
link |
I actually like that I have a sense of humility of like,
link |
I may not found another thing that's nearly as big
link |
so I have to work twice as hard
link |
or I have to like learn twice as much.
link |
I have to, we haven't talked about machine learning yet,
link |
but my favorite thing is all these like famous,
link |
you know, tech guys who have worked in the industry,
link |
pontificating about the future of machine learning
link |
and how it's gonna kill us all, right?
link |
And like, I'm pretty sure they've never tried
link |
to build anything with machine learning themselves.
link |
Yes, so there's a nice line between people
link |
that actually build stuff with machine,
link |
like actually program something
link |
or at least understand some of those fundamentals
link |
and the people that are just saying philosophical stuff
link |
or journalists and so on.
link |
It's an interesting line to walk
link |
because the people who program are often not philosophers.
link |
Or don't have the attention,
link |
they can't write an op ed for the Wall Street Journal,
link |
like it doesn't work.
link |
So like, it's nice to be both a little bit,
link |
like to have elements of both.
link |
My point is the fact that I have to learn stuff from scratch
link |
or that I choose to are like.
link |
Yeah, I mean, again, I have a lot of advantages.
link |
I like, but my point is it's awesome
link |
to be back in a game where you have to fight.
link |
That is, that's fun.
link |
So being humble, being vulnerable,
link |
it's an important aspect of a leader
link |
and I hope it serves me well,
link |
but like, I can't fast forward 10 years to now.
link |
I've just, that's my game plan.
link |
Before I forget, I have to ask you one last thing
link |
What do you think about the whistleblower,
link |
Frances Haugen, recently coming out
link |
and saying that Facebook is aware of Instagram's
link |
harmful effect on teenage girls
link |
as per their own internal research studies on the matter?
link |
What do you think about this baby of yours,
link |
Instagram being under fire now,
link |
as we've been talking about under the leadership of Facebook?
link |
You know, I often question, where does the blame lie?
link |
Is the blame at the people that originated the network me?
link |
Is the blame at like the decision to combine the network
link |
with another network with a certain set of values?
link |
Is the blame at how it gets run after I left?
link |
Like, is it the driver or is it the car, right?
link |
Is it that someone enabled these devices in the first place?
link |
If you go to an extreme, right?
link |
Or is it the users themselves, just human nature?
link |
Is it just the way of human nature?
link |
Sure, and like the idea that we're gonna find
link |
a mutually exclusive answer here is crazy.
link |
There's not one place that's a combination
link |
of a lot of these things.
link |
And then the question is like, is it true at all, right?
link |
Like I'm not actually saying that's not true
link |
or that it's true, but there's always more nuance here.
link |
Do I believe that social media has an effect
link |
Well, it's got it, they use it a lot.
link |
And I bet you there are a lot of positive effects
link |
and I bet you there are negative effects,
link |
just like any technology.
link |
And where I've come to in my thinking on this
link |
is that I think any technology has negative side effects.
link |
The question is, as a leader, what do you do about them?
link |
And are you actively working on them
link |
or do you just like not really believe in them?
link |
If you're a leader that sits there and says,
link |
well, we're gonna put an enormous amount
link |
of resources against this.
link |
We're gonna acknowledge when there are true criticisms,
link |
we're gonna be vulnerable and that we're not perfect
link |
and we're gonna go fix them
link |
and we're gonna be held accountable along the way.
link |
I think that people generally really respect that.
link |
But I think that where Facebook I think has had issues
link |
in the past is where they say things like,
link |
can't remember what Mark said about misinformation
link |
during the election.
link |
There was that like famous quote where he was like,
link |
it's pretty crazy to think that Facebook had anything
link |
to do with this election.
link |
Like that was something like that quote.
link |
And I don't remember what stage he was on and yeah.
link |
But ooh, that did not age well, right?
link |
Like you have to be willing to say,
link |
well, maybe there's something there and wow,
link |
like I wanna go look into it
link |
and truly believe it in your gut.
link |
But if people look at you and how you act
link |
and what you say and don't believe you truly feel that way.
link |
It's not just the words you say, but how you say them
link |
and that people believe they actually feel the pain
link |
of having caused any suffering in the world.
link |
So to me, it's much more about your actions
link |
and your posture post event
link |
than it is about debugging the why.
link |
Cause I don't know, is it like, I don't know this research.
link |
It was written well after I left, right?
link |
Like, is it the algorithm?
link |
Is it the explore page?
link |
Is it the people you might know unit connecting you
link |
to ideas that are dangerous?
link |
Like I really don't know.
link |
So we'd have to have a much deeper,
link |
I think dive to understand where the blame lies.
link |
What's very unpleasant to me to consider,
link |
now, I don't know if this is true,
link |
but to consider the very fact that there might be
link |
some complicated games being played here.
link |
For example, as somebody, I really love psychology
link |
and I love it enough to know that the field
link |
is pretty broken in the following way.
link |
It's very difficult to study human beings well at scale
link |
because the questions you ask affect the results.
link |
You can basically get any results you want.
link |
And so you have an internal Facebook study
link |
that asks some question of which we don't know
link |
the full details and there's some kind of analysis,
link |
but that's just the one little tiny slice
link |
into some much bigger picture.
link |
And so you can have thousands of employees at Facebook.
link |
One of them comes out and picks whatever narrative,
link |
knowing that they become famous,
link |
coupled with the other really uncomfortable thing
link |
I see in the world, which is journalists seem to understand
link |
they get a lot of clickbait attention
link |
from saying something negative about social networks.
link |
Certain companies, like they even get some clickbait stuff
link |
about Tesla or about, especially when it's like,
link |
when there's a public famous CEO type of person,
link |
if they get a lot of views on the negative, not the positive,
link |
the positive, they'll get, I mean,
link |
it actually goes to the thing you were saying before,
link |
if there's a hot, sexy new product,
link |
that's great to look forward to, they get positive on that,
link |
but absent a product, it's nice to have like the CEO
link |
messing up in some kind of way.
link |
And so couple that with the whistleblower
link |
and with this whole dynamic of journalism and so on,
link |
you know, with Social Dilemma being really popular,
link |
documentary, it's like, all right,
link |
my concern is there's deep flaws in human nature here
link |
in terms of things we need to deal with,
link |
like the nature of hate, of bullying,
link |
all those kinds of things.
link |
And then there's people who are trying to use that
link |
potentially to become famous and make money
link |
off of blaming others for causing more of the problem
link |
as opposed to helping solve the problem.
link |
So I don't know what to think.
link |
I'm not saying this is like, I'm just uncomfortable
link |
with, I guess, not knowing what to think about any of this
link |
because a bunch of folks I know that work at Facebook
link |
on the machine learning side, so Yann LeCun,
link |
I mean, they're quite upset about what's happening
link |
because there's a lot of really brilliant,
link |
good people inside Facebook that are trying to do good.
link |
And so like all of this press, Yann is one of them,
link |
and he has an amazing team
link |
with the machine learning researchers.
link |
Like he's really upset with the fact
link |
that people don't seem to understand
link |
that the portrayal does not represent
link |
the full nature of efforts that's going on at Facebook.
link |
So I don't know what to think about that.
link |
Well, you just, I think, very helpfully explained
link |
the nuance of the situation
link |
and why it's so hard to understand.
link |
But a couple of things.
link |
One is I think I have been surprised
link |
at the scale with which some product manager
link |
can do an enormous amount of harm
link |
to a very, very large company
link |
by releasing a trove of documents.
link |
Like I think I read a couple of them when they got published
link |
and I haven't even spent any time going deep.
link |
Part of it's like, I don't really feel like reliving
link |
a previous life, but wow.
link |
Like talk about challenging the idea of open culture
link |
and like what that does to Facebook internally.
link |
If Facebook was built, like I remember like my office,
link |
we had this like no visitors rule around my office
link |
because we always had like confidential stuff up
link |
on the walls and everyone was super angry
link |
because they're like, that goes against our culture
link |
of transparency and like marks in the fish cube
link |
or whatever they call it, the aquarium,
link |
I think they called it, where like literally anyone could see
link |
what he was doing at any point and I don't know.
link |
I mean, other companies like Apple have been quiet
link |
slash locked down, Snapchat's the same way for a reason.
link |
And I don't know what this does to transparency
link |
on the inside of startups that value that.
link |
I think that it's a seminal moment and you can say,
link |
well, you should have nothing to hide, right?
link |
But to your point, you can pick out documents
link |
that show anything, right?
link |
What happens to transparency inside of startups
link |
and the culture that startups or companies
link |
in the future will grow, like the startup of the future
link |
that becomes the next Facebook will be locked down
link |
and what does that do, right?
link |
So that's part one.
link |
Part two, like I don't think that you could design
link |
a more like a well orchestrated handful of events
link |
from the like 16 minutes to releasing the documents
link |
in the way that they were released at the right time.
link |
That takes a lot of planning and partnership.
link |
And it seems like she has a partner at some firm, right?
link |
That probably helped a lot with this, but man,
link |
at a personal level, if you're her,
link |
you'd have to really believe in what you are doing,
link |
really believe in it because you are personally
link |
putting your ass on the line, right?
link |
Like you've got a very large company
link |
that doesn't like enemies, right?
link |
It takes a lot of guts and I don't love
link |
these conspiracy theories about like,
link |
oh, she's being financed from some person or people.
link |
Like I don't love them because that's
link |
like the easy thing to say.
link |
I think the Occam's razor here is like someone thought
link |
they were doing something wrong
link |
and was like very, very courageous.
link |
And I don't know if courageous is the word,
link |
but like, so without getting into like,
link |
Is she courageous?
link |
Like, let's put that aside for a second.
link |
Then there are the documents themselves.
link |
They say what they say.
link |
To your point, a lot of the things that like people
link |
have been worried about already in the documents
link |
or they're already been said externally.
link |
And I don't know, I'm just like, I'm thankful
link |
that I am focused on new things with my life.
link |
Well, let me just say, I just think it's a really
link |
hard problem that probably Facebook and Twitter
link |
are trying to solve.
link |
I'm actually just fascinated by how hard this problem is.
link |
There are fundamental issues at Facebook in tone
link |
and in an approach of how product gets built
link |
and the objective functions.
link |
And since people, organizations are not people.
link |
So yawn and fair, right?
link |
Like there are a lot of really great people
link |
who like literally just want to push
link |
reinforcement learning forward.
link |
They literally just want to teach a robot
link |
to touch, feel, lift, right?
link |
Like they're not thinking about political misinformation,
link |
But there's a strong connection between what funds
link |
that research and an enormously profitable machine
link |
that has trade offs.
link |
And one cannot separate the two.
link |
You are not completely separate from the system.
link |
So I agree, it can feel really frustrating
link |
to feel if you're internal there,
link |
that you're working on something completely unrelated
link |
and you feel like your group's good.
link |
I can understand that.
link |
But there's some responsibility still.
link |
You have to acknowledge, it's like the Ray Dalio thing.
link |
You have to look in the mirror and see if there's problems
link |
and you have to fix those problems.
link |
You mentioned machine learning reinforcement quite a bit.
link |
I mean, to me, social networks is one of the exciting places,
link |
recommender systems where machine learning is applied.
link |
Where else in the world, in the space of possibilities
link |
over the next five, 10, 20 years,
link |
do you think we're going to see impact of machine learning
link |
when you try, on the philosophical level,
link |
on a technical level, what do you think?
link |
Or within social networks themselves?
link |
Well, I think the obvious answers are climate change.
link |
Think about how much fuel
link |
or just waste there is in energy consumption today
link |
because we don't plan accordingly,
link |
because we take the least efficient route or...
link |
The logistics and stuff, the supply chain,
link |
all that kind of stuff.
link |
Yeah, I mean, listen, if we're gonna fight climate change,
link |
like one really way, one awesome way to do it
link |
is figure out how to optimize how we operate as a species
link |
and minimize the amount of energy we consume
link |
to maximize whatever economic impact we wanna have.
link |
Because right now those two are very much tied together.
link |
And I don't believe that that has to be the case.
link |
There's this really interesting, you've read it.
link |
For people who are listening,
link |
there's this really interesting paper
link |
on reinforcement learning
link |
and energy consumption inside buildings.
link |
It's like one of the seminal ones, right?
link |
But imagine that at massive scale.
link |
That's super interesting.
link |
I mean, they've done resource planning for servers
link |
for peak load using reinforcement learning.
link |
I don't know if that was at Google or somewhere else,
link |
but like, okay, great, you do it for servers,
link |
but what if you could do it for just capacity
link |
and general energy capacity for cities
link |
and planning for traffic?
link |
And of course there's all the self driving cars
link |
and I don't know, like I'm not gonna pontificate
link |
on like crazy ideas using reinforcement learning
link |
or machine learning.
link |
It's just so clear to me
link |
that humans don't think quickly enough.
link |
So it's interesting to think about machine learning
link |
helping a little bit at scale.
link |
So a little bit to a large number of people
link |
that has a huge impact.
link |
So if you optimize, say Google Maps, something like that,
link |
trajectory planning or what a map quest first.
link |
Getting here, I looked and it was like,
link |
here's the most energy efficient route.
link |
And I was like, I'm gonna be late.
link |
I need to take the fastest route.
link |
As opposed to unrolling the map.
link |
Like, and then that's going to be very inefficient
link |
I was definitely the other day,
link |
like part of the Epsilon of Epsilon Greedy
link |
with Waze where like I was sent on like a weird route
link |
that I could tell they're like,
link |
we just need to collect data at this road.
link |
Like we just, Kevin's.
link |
You were the ant they sent out for exploration.
link |
Kevin's definitely gonna be the guinea pig.
link |
And great, now we have.
link |
Did you at least feel pride?
link |
Oh, going through it, I was like, oh, this is fun.
link |
Like now they get data about this weird shortcut.
link |
And actually I hit all the green lights networked.
link |
I'm like, this is a problem.
link |
Bad data, they're just gonna imagine.
link |
I could see you slowing down and stopping at a green light
link |
just to give them the right kind of data.
link |
But to answer your question,
link |
like I feel like that was fairly unsatisfying
link |
and it's easy to say climate change.
link |
But what I would say is at Instagram,
link |
everything we applied machine learning to
link |
got better for users and it got better for the company.
link |
I didn't fully understand it as an executive.
link |
And I think that's actually one of the issues
link |
that, and when I say understand,
link |
I mean the mathematics of it.
link |
Like I understand what it does.
link |
I understand that it helps.
link |
But there are a lot of executives now
link |
that talk about it in the way
link |
that they talk about the internet
link |
or they talked about the internet like 10 years ago.
link |
They're like, we're gonna build mobile.
link |
And you're like, what does that mean?
link |
They're like, we're just gonna do mobile.
link |
And you're like, okay.
link |
So my sense is the next generation of leaders
link |
will have grown up having had classes
link |
in reinforcement learning, supervised learning, whatever.
link |
And they will be able to thoughtfully apply it
link |
to their companies and the places that it is needed most.
link |
And that's really cool.
link |
Cause I mean, talk about efficiency gains.
link |
That's what excites me the most about it.
link |
Yeah, so there's, it's interesting just to get a fundamental
link |
first principles understanding
link |
of certain concepts of machine learning.
link |
So supervised learning from an executive perspective,
link |
supervised learning, you have to have a lot of humans
link |
label a lot of data.
link |
So the question there is, okay,
link |
can we gather a large amount of data
link |
that can be labeled well?
link |
And that's the question Tesla asked,
link |
like can we create a data engine
link |
that keeps sending an imperfect machine learning system
link |
out there, whenever it fails, it gives us data back.
link |
We label it by human and we send it back and forth
link |
Then there's Yann LeCun's excited
link |
about the self supervised learning
link |
where you do much less human labeling
link |
and there's some kind of mechanism for the system
link |
to learn it by itself on the human generated data.
link |
And then there's the reinforcement learning,
link |
which is like basically allowing,
link |
it's applying the alpha zero technology
link |
that allow through self play to learn how to solve
link |
the game of Go and achieve incredible levels
link |
at the game of chess.
link |
Can you formulate the problem you're trying to solve
link |
in a way that's amenable to reinforcement learning?
link |
And can you get the right kind of signal at scale?
link |
Cause you need a lot of signal.
link |
And that's kind of fascinating to see which part
link |
of a social network can you convert
link |
into reinforcement learning problem.
link |
The fascinating thing about reinforcement learning,
link |
I think, is that we now have learned
link |
to apply neural networks to guess the Q function,
link |
basically the values for any state in action.
link |
And that is fascinating cause we used to just like,
link |
I don't know, have like a linear regression,
link |
like hope it worked and that was the fanciest version of it.
link |
But now you look at it, I'm like trying to learn this stuff
link |
and I look at it and I'm like,
link |
there are like 17 different acronyms of different ways
link |
you can try to apply this.
link |
No one quite agrees, like what's the best.
link |
Generally, if you're trying to like build a neural network,
link |
there are pretty well trodden ways of doing that.
link |
You use Adam, you use ReLU,
link |
like there's just like general good ideas.
link |
And in reinforcement learning,
link |
I feel like the consensus is like, it totally depends.
link |
And by the way, it's really hard to get it to converge
link |
and it's noisy and it like,
link |
so there are all these really interesting ideas
link |
around building simulators.
link |
You know, like for instance, in self driving, right?
link |
Like you don't want to like actually have someone
link |
getting in an accident to learn that an accident is bad.
link |
So you start simulating accidents,
link |
simulating aggressive drivers,
link |
just simulating crazy dogs that run into the street and,
link |
wow, fascinating, right?
link |
Like my mind starts racing and then the question is,
link |
okay, forget about self driving cars.
link |
Let's talk about social networks.
link |
How can you produce a better, more thoughtful experience
link |
using these types of algorithms?
link |
And honestly, in talking to some of the people
link |
that work at Facebook and old Instagrammers,
link |
most people are like, yeah, we tried a lot of things,
link |
didn't quite ever make it work.
link |
I mean, for the longest time,
link |
Facebook ads was effectively a logistic regression, okay?
link |
I don't know what it is now,
link |
but like if you look at this paper
link |
that they published back in the day,
link |
it was literally just a logistic regression.
link |
Made a lot of money.
link |
So even at these like extremely large scales,
link |
if we are not yet touching
link |
what reinforcement learning can truly do,
link |
imagine what the next 10 years looks like.
link |
So I really liked the use of reinforcement learning
link |
as part of the simulation, for example,
link |
like with self driving cars, it's modeling pedestrians.
link |
So the nice thing about reinforcement learning,
link |
it can be used to learn agents within the world.
link |
So they can learn to behave properly.
link |
Like you can teach pedestrians to,
link |
like you don't hard code the way they behave,
link |
they learn how to behave.
link |
In that same way, I do have a hope,
link |
was it Jack Dorsey talks about healthy conversations.
link |
You talked about meaningful interactions, I believe.
link |
Like simulating interactions.
link |
So you can learn how to manage that, it's fascinating.
link |
So where most of your algorithm development happens
link |
in virtual worlds, and then you can really learn
link |
how to design the interface,
link |
how you design a bunch of aspects of the experience
link |
in terms of how you select what's shown in the feed,
link |
all those kinds of things.
link |
It feels like if you can connect reinforcement learning
link |
to that, that's super exciting.
link |
Yep, and I think if you have a company and leadership
link |
that believe in doing the right things
link |
and can apply this technology in the right way,
link |
some really special stuff can happen.
link |
It is mostly likely going to be a group of people
link |
we've never heard about, start up from scratch, right?
link |
And you asked if like new social networks could be built,
link |
I've got to imagine they will be.
link |
And whoever starts it, it might be some kids in a garage
link |
that took these classes from these people, you, right?
link |
And they're building all of these things
link |
with this tech at the core.
link |
So I'm trying not to be someone who just like throws
link |
around reinforcement learning as a buzzword.
link |
I truly believe that it is the most cutting edge
link |
in what can happen in social networks.
link |
And I also believe it's super hard.
link |
Like it's super hard to make it work.
link |
It's super hard to do it at scale.
link |
It's super hard to find people that truly understand it.
link |
So I'm not gonna say that like,
link |
I think it'll be applied in social networks
link |
before we have true self driving.
link |
Yeah, we could argue about this for a long time,
link |
but yes, I agree with you.
link |
I think self driving is way harder than people realize.
link |
Let me ask you in terms of that kid in the garage
link |
or those couple of kids in the garage,
link |
what advice would you give to them
link |
if they wanna start a new social network or a business?
link |
What advice would you give to somebody
link |
with a big dream and a young startup?
link |
To me, you have to choose to do something
link |
that even if it fails, like it was so fun, right?
link |
Like we never started Instagram knowing
link |
it was going to be big.
link |
We started Instagram because we loved photography.
link |
We loved social networks.
link |
I had seen what other social networks had done
link |
and I thought, hmm, maybe we did a spin on this,
link |
but like nowhere was our feet predestined.
link |
Like it wasn't like, it wasn't written out anywhere
link |
that everything was gonna go great.
link |
And I often think the counterfactual,
link |
like what if it had not gone well?
link |
I would have been like, I don't know, that was fun.
link |
We raised some money, we learned some stuff
link |
and does it position you well for the next experience?
link |
That's the advice that I would give
link |
to anyone wanting to start something today,
link |
which is like, does this meet with your ultimate goals?
link |
Not wealth, not fame, none of that,
link |
because all of that, by the way, is bullshit.
link |
Like you can get super famous and super wealthy.
link |
And I think generally those are not things that,
link |
again, it's easy to say with like a lot of money
link |
that somehow like it's not good to have a lot of money.
link |
It's just, I think that complicates life enormously
link |
in a way that people don't fully comprehend.
link |
So I think it is way more interesting to shoot for,
link |
can I make something that people love,
link |
that provides value in the world,
link |
that I love building, that I love working on, right?
link |
That's what I would do if I were starting from scratch.
link |
And by the way, like in some ways
link |
that I will do that personally,
link |
which is like choose the thing
link |
that you get up every morning and you're like,
link |
I love this, even when it's painful.
link |
Even when it's painful.
link |
What about a social network specifically?
link |
If you were to imagine, put yourself in the mind of some.
link |
I can't compete against myself.
link |
I can't give out ideas.
link |
No, but it's like high level.
link |
You can focus on community.
link |
I said that as a half joke.
link |
In all honesty, I think these things are so hard to build
link |
that like ideas are a dime a dozen, but.
link |
You have talked about keeping it simple.
link |
Which is a liberating idea.
link |
My model is it's three circles and they overlap.
link |
One circle is what do I have experience at?
link |
Slash, what am I good at?
link |
I don't like saying what am I good at
link |
because it just like seems like,
link |
what do I have experience in, right?
link |
What can I bring to the table?
link |
What am I excited about is the other circle.
link |
What's just super cool, right?
link |
That I want to work on because even when this is hard,
link |
I think it's so cool.
link |
I want to stick with it.
link |
And the last circle is like, what does the world need?
link |
And if that circle ain't there,
link |
it doesn't matter what you work on.
link |
Cause there are a lot of startups that exist
link |
that just no one needs or very small markets need.
link |
But if you want to be successful,
link |
I think if you're like, if you're good at it,
link |
you have, sorry, if you're good at it,
link |
you're passionate about it and the world needs it.
link |
I mean, this sounds simple,
link |
not enough people sit down and just think
link |
about those circles and think, do these things overlap?
link |
And then can I get that middle section?
link |
It's small, but can I get that middle section?
link |
I think a lot about that personally.
link |
And then you have to be really honest about the circle
link |
that you're good at and really honest about the circle
link |
that the world needs.
link |
And as opposed to really honest about the passion,
link |
like what do you actually love?
link |
As opposed to like some kind of dream of making money,
link |
all those kinds of stuff, like literally love doing.
link |
I had a former engineer who decided to start a startup
link |
and I was like, are you sure you want to start a company
link |
versus like join something else?
link |
Because being a coach of an NBA team and playing basketball
link |
are two very, very different things.
link |
And like not everyone fully understands the difference.
link |
I think you can kind of do it both.
link |
And I don't know, jury's out on that one
link |
because like they're in the middle of it now.
link |
But it's really important to figure out
link |
what you're good at, not be full of yourself,
link |
like truly look at your track record.
link |
What's the saying like, it ain't bragging if you can do it.
link |
But too many people are delusional
link |
and like think they're better at things
link |
than they actually are,
link |
or think there's a bigger market than there actually is.
link |
When you confuse your passion for things with a big market,
link |
that's really scary, right?
link |
Like just because you think it's cool
link |
doesn't mean that it's a big business opportunity.
link |
So like, what evidence do you have?
link |
Again, I'm a fairly like, I'm a strict rationalist on this.
link |
And like sometimes people don't like working with me
link |
because I'm pretty pragmatic about things.
link |
Like I'm not Elon, like I don't sit
link |
and make bold proclamations about visiting Mars.
link |
Like that's just not how I work.
link |
I'm like, okay, I want to build this really cool thing
link |
that's fairly practical and I think we could do it.
link |
And it's in this way.
link |
And what's cool though is like, that's just my sweet spot.
link |
I'm not like, I just, I can't with a straight face
link |
talk about the metaphors.
link |
I can't, I just, it's not me.
link |
What do you think about the Facebook renaming itself to?
link |
I didn't mean that as a dig.
link |
I just literally mean like, I'm fairly,
link |
I like to live in the next five years.
link |
And like, what things can I get out in a year
link |
that people will use at scale?
link |
And so it's just, again, those circles I think are different
link |
for different people, but it's important to realize
link |
that like market matters, you being good at it matters
link |
and having passion for it matters.
link |
Your question, sorry.
link |
Well, on that last, on this topic in terms of funding,
link |
is there, by way of advice,
link |
was funding in your own journey helpful, unhelpful?
link |
Like is there a right time to get funding, venture funding
link |
or anything, borrow some money from your parents?
link |
Like is money getting in the way?
link |
Is the timing important?
link |
Is there some kind of wisdom you can give there
link |
because you were exceptionally successful very quickly?
link |
Funding helps as long as it's from the right people.
link |
That includes yourself.
link |
And I'll talk about myself funding myself in a second,
link |
which is like, because I can fund myself
link |
doing whatever projects I can do,
link |
I don't really have another person putting pressure on me
link |
except for myself and that creates strange dynamics, right?
link |
But let's like talk about people getting funding
link |
from a venture capitalist initially.
link |
We raised money from Matt Kohler at Benchmark.
link |
He's brilliant, amazing guy, very thoughtful.
link |
And he was very helpful early on.
link |
But I have stories from entrepreneurs
link |
where they raised money from the wrong person
link |
or the wrong firm where incentives weren't aligned.
link |
They didn't think in the same way
link |
and bad things happened because of that.
link |
The boardroom was always noisy.
link |
There were fights, like we just never had that.
link |
I think like capital these days
link |
is kind of a dime a dozen, right?
link |
Like as long as you're fundable,
link |
like it seems like there's money out there
link |
is what I'm hearing.
link |
It's really important that you are aligned
link |
and that you think of raising money
link |
as hiring someone for your team rather than taking money
link |
if capital is plentiful, right?
link |
It provides a certain amount of pressure
link |
to do the right thing that I think is healthy
link |
And it keeps you real and honest
link |
because they don't wanna lose their money.
link |
They're paid to not lose their money.
link |
The problem, maybe I could depersonalize it,
link |
but like I remember having lunch with Elon.
link |
It's only happened once.
link |
And I asked him, like I was trying to figure out
link |
what I was doing after Instagram, right?
link |
And I asked him something about like angel investing.
link |
And he looked at me with a straight face.
link |
He was like, why the F would I do that?
link |
I was like, I don't know.
link |
Like you're connected.
link |
Like seems like he's like, I only invest in myself.
link |
I was like, Ooh, okay.
link |
You know, like not the confidence.
link |
I was just like, what a novel idea.
link |
It's like, yeah, if you have money,
link |
like why not just put it against your bag
link |
and like enable you're visiting Mars or something, right?
link |
Like that's awesome, great.
link |
But I had never really thought of it that way.
link |
But also with that comes an interesting dynamic
link |
where you don't actually have people
link |
who are gonna lose that money telling you,
link |
hey, don't do this or, hey, you need to face this reality.
link |
So you need to create other versions of that truth teller.
link |
And whatever I do next,
link |
that's gonna be one of the interesting challenges
link |
is how do you create that truth telling situation?
link |
And that's part of why, by the way,
link |
I think someone like Jack, when you start Square,
link |
you have money, but you still, you bring on partners
link |
because I think it creates
link |
a truth telling type environment.
link |
I'm still trying to figure this out.
link |
Like it's an interesting dynamic.
link |
So you're thinking of perhaps launching some kind of venture
link |
where you're investing in yourself?
link |
I mean, I'm 37 going on 38 next month.
link |
I have a long life to live.
link |
I'm definitely not gonna sit on the beach, right?
link |
So I'm gonna do something at some point
link |
and I gotta imagine I will like help fund it, right?
link |
So the other way of thinking about this
link |
is you can park your money in the SMP,
link |
because the SMP has done wonderfully well last year, right?
link |
Or you can invest in yourself.
link |
And if you're not gonna invest in yourself,
link |
you probably shouldn't do a startup.
link |
It's kind of the way of thinking about it.
link |
And you can invest in yourself in the way Elon does,
link |
which is basically go all in on this investment.
link |
Maybe that's one way to achieve accountability
link |
is like you're kind of screwed if you fail.
link |
Yeah, that's, yeah.
link |
I personally like that.
link |
I like burning bridges behind me
link |
so that I'm fucked if it fails.
link |
It's really important though.
link |
One of the things I think Mark said to me early on
link |
that sticks with me that I think is true.
link |
We were talking about people
link |
who had left like operating roles
link |
and started doing venture or something.
link |
He was like, a lot of people convince themselves
link |
they work really hard.
link |
Like they think they work really hard
link |
and they put on the show
link |
and in their minds they work really hard,
link |
but they don't work very hard.
link |
There is something about lighting a fire underneath you
link |
and burning bridges such that you can't turn back.
link |
That I think, we didn't talk about this specifically,
link |
but I think you're right.
link |
There is, you need to have that
link |
because there's the self delusion at a certain scale.
link |
Oh, I have so many board calls.
link |
Oh, like we have all these things to figure out.
link |
It's like, this is one of the hard parts
link |
about it being an operator.
link |
It's like, there are so many people
link |
that have made a lot of money not operating,
link |
but operating is just one of the hardest things on earth.
link |
It is just so effing hard.
link |
It is, you're dealing with real humans,
link |
not just like throwing capital in and hoping it grows.
link |
I'm not undermining the VC mindset.
link |
I think it's a wonderful thing and needed
link |
and so many wonderful VCs I've worked with.
link |
But yeah, like when your ass is on the line
link |
and it's your money, it's...
link |
Talk to me in 10 years, we'll see how it goes.
link |
Yeah, but like you were saying, that is a source.
link |
When you wake up in the morning
link |
and you look forward to the day full of challenges,
link |
that's also where you can find happiness.
link |
Let me ask you about love and friendship.
link |
What's the role in this heck of a difficult journey
link |
you have been on of love, of friendship?
link |
What's the role of love in the human condition?
link |
Well, first things first,
link |
the woman I married, my wife, Nicole,
link |
there's no way I could do what I do if we weren't together.
link |
She had the filter idea.
link |
Yeah, yeah, exactly.
link |
We didn't go over that story.
link |
Everything is a partnership, right?
link |
And to achieve great things,
link |
it's not about like someone pulling their weight in places.
link |
Like it's not like someone's supporting you
link |
so that you could do this other thing.
link |
It's literally like,
link |
Mike and I and our partnership as cofounders is fascinating
link |
because I don't think Instagram would have happened
link |
without that partnership.
link |
Like either him or me alone, no way.
link |
We pushed and pulled each other in a way
link |
that allowed us to build a better thing because of it.
link |
Nicole, she pushed me to work on the filters early on.
link |
And yes, that's exciting.
link |
It's a fun story, right?
link |
But the truth of it is being able to like level
link |
with someone about how hard the process is
link |
and have someone see you for who you are before Instagram
link |
and know that there's a constant you throughout all of this
link |
and be able to call you when you're drifting from that,
link |
but also support you when you're trying to stick with that.
link |
That's, I mean, that's true friendship slash love,
link |
whatever you want to call it.
link |
But also it was for someone not to care.
link |
I remember Nicole saying,
link |
hey, like I know you're going to do this Instagram thing.
link |
You should, I guess it was bourbon at the time.
link |
You should do it because, you know,
link |
even if it doesn't work,
link |
we can move to like a smaller apartment and it'll be fine.
link |
Like we'll make it work.
link |
How beautiful is that, right?
link |
That's almost like a superpower
link |
that gives you permission to fail.
link |
And somehow that actually leads to success.
link |
But also she's like the least impressed
link |
about Instagram of anyone.
link |
She's like, yeah, it's great.
link |
But like, I love you for you.
link |
Like, I like that you're like a decent cook.
link |
That's beautiful with the Gantt chart and Thanksgiving,
link |
which I still think is a brilliant effing idea.
link |
Big, ridiculous question.
link |
Have you, you're old and wise at this stage.
link |
So have you discovered meaning to this whole thing?
link |
Why the hell are we descendants of apes here on earth?
link |
What's the meaning of it?
link |
What's the meaning of life?
link |
And I am, so the crazy,
link |
so the best learning for me has been like,
link |
no matter what level of success you achieve,
link |
you're still worried about similar things,
link |
just maybe on a slightly different scale.
link |
You're still concerned about the same thing.
link |
You're still self conscious about the same things.
link |
Just like, and actually that moment going through that
link |
is what makes you believe there's gotta be like
link |
more machinery to life or purpose to life.
link |
And that we're all chasing these materialistic things,
link |
but you start realizing like,
link |
it's almost like, you know, the Truman Show
link |
when he gets to the edge and he like knocks against it.
link |
Like there's this awakening that happens
link |
when you get to that edge that you realize,
link |
oh, like sure, it's great.
link |
It's great that we all chase money and fame and success.
link |
But you hit the edge and I'm not even claiming
link |
I hit an edge like Elon's hit an edge.
link |
Like there's clearly larger scales.
link |
But what's cool is you learn that,
link |
like it doesn't actually matter
link |
and that there are all these other things that truly matter.
link |
That's not a case for working less hard.
link |
That's not a case for taking it easy.
link |
That's not a case for the four day work week.
link |
What that is a case for is designing your life
link |
exactly the way you want to design it.
link |
Cause I don't know, I think we go around the earth,
link |
you know, the sun a certain number of times
link |
and then we die and then that's it.
link |
Are you afraid of that moment?
link |
In fact, or at least not yet.
link |
Listen, I'm like a pilot, like I do crazy things
link |
and I like, no, I like, if anything, I'm like,
link |
oh, I got to choose mindfully and purposefully
link |
the thing I am doing right now and not just fall into it
link |
because you're going to wake up one day and ask yourself
link |
why the hell you spent the last 10 years doing X, Y or Z?
link |
So I guess my like shorter answer to this is
link |
doing things on purpose because you choose to do them.
link |
So important in life and not just like floating
link |
down the river of life, hitting branches along the way
link |
cause you will hit branches, right?
link |
But rather like literally plotting a course
link |
and not having a 10 year plan,
link |
but just choosing every day to opt in.
link |
That I think has been more like,
link |
I haven't figured out the meaning of life
link |
by any stretch of the imagination,
link |
but it certainly isn't money and it certainly isn't fame
link |
and it certainly isn't travel.
link |
And it's like, and it's way more of like opting
link |
into the game you love playing.
link |
Every day opting in.
link |
Just opting in and like, don't let it happen.
link |
Kevin, it's great to end on love and the meaning of life.
link |
This was an amazing conversation.
link |
It was a lot of fun, thank you.
link |
You gave me like a light into some fascinating aspects
link |
of this technical world.
link |
And I can't honestly wait to see what you do next.
link |
Thank you so much.
link |
Thanks for having me.
link |
Thanks for listening to this conversation
link |
with Kevin Systrom.
link |
To support this podcast, please check out our sponsors
link |
in the description.
link |
And now let me leave you with some words
link |
from Kevin Systrom himself.
link |
Focusing on one thing and doing it really, really well
link |
can get you very far.
link |
Thank you for listening and hope to see you next time.