back to indexErik Brynjolfsson: Economics of AI, Social Networks, and Technology | Lex Fridman Podcast #141
link |
The following is a conversation with Erik Brynjolfsson.
link |
He's an economics professor at Stanford
link |
and the director of Stanford's Digital Economy Lab.
link |
Previously, he was a long, long time professor at MIT
link |
where he did groundbreaking work
link |
on the economics of information.
link |
He's the author of many books,
link |
including The Second Machine Age
link |
and Machine Platform Crowd,
link |
coauthored with Andrew McAfee.
link |
Quick mention of each sponsor,
link |
followed by some thoughts related to the episode.
link |
Ventura Watches, the maker of classy,
link |
well performing watches.
link |
Four Sigmatic, the maker of delicious mushroom coffee.
link |
ExpressVPN, the VPN I've used for many years
link |
to protect my privacy on the internet.
link |
And CashApp, the app I use to send money to friends.
link |
Please check out these sponsors in the description
link |
to get a discount and to support this podcast.
link |
As a side note, let me say that the impact
link |
of artificial intelligence and automation
link |
on our economy and our world
link |
is something worth thinking deeply about.
link |
Like with many topics that are linked
link |
to predicting the future evolution of technology,
link |
it is often too easy to fall into one of two camps.
link |
The fear mongering camp
link |
or the technological utopianism camp.
link |
As always, the future will land us somewhere in between.
link |
I prefer to wear two hats in these discussions
link |
and alternate between them often.
link |
The hat of a pragmatic engineer
link |
and the hat of a futurist.
link |
This is probably a good time to mention Andrew Yang,
link |
the presidential candidate who has been
link |
one of the high profile thinkers on this topic.
link |
And I'm sure I will speak with him
link |
on this podcast eventually.
link |
A conversation with Andrew has been on the table many times.
link |
Our schedules just haven't aligned,
link |
especially because I have a strongly held to preference
link |
for long form, two, three, four hours or more,
link |
I work hard to not compromise on this.
link |
Trust me, it's not easy.
link |
Even more so in the times of COVID,
link |
which requires getting tested nonstop,
link |
staying isolated and doing a lot of costly
link |
and uncomfortable things that minimize risk for the guest.
link |
The reason I do this is because to me,
link |
something is lost in remote conversation.
link |
That something, that magic,
link |
I think is worth the effort,
link |
even if it ultimately leads to a failed conversation.
link |
This is how I approach life,
link |
treasuring the possibility of a rare moment of magic.
link |
I'm willing to go to the ends of the world
link |
for just such a moment.
link |
If you enjoy this thing, subscribe on YouTube,
link |
review it with five stars on Apple Podcast,
link |
follow on Spotify, support on Patreon,
link |
connect with me on Twitter at Lex Friedman.
link |
And now here's my conversation with Erik Brynjolfsson.
link |
You posted a quote on Twitter by Albert Bartlett
link |
saying that the greatest shortcoming of the human race
link |
is our inability to understand the exponential function.
link |
Why would you say the exponential growth
link |
is important to understand?
link |
Yeah, that quote, I remember posting that.
link |
It's actually a reprise of something Andy McAfee and I said
link |
in the second machine age,
link |
but I posted it in early March
link |
when COVID was really just beginning to take off
link |
and I was really scared.
link |
There were actually only a couple dozen cases,
link |
maybe less at that time,
link |
but they were doubling every like two or three days
link |
and I could see, oh my God, this is gonna be a catastrophe
link |
and it's gonna happen soon,
link |
but nobody was taking it very seriously
link |
or not a lot of people were taking it very seriously.
link |
In fact, I remember I did my last in person conference
link |
that week, I was flying back from Las Vegas
link |
and I was the only person on the plane wearing a mask
link |
and the flight attendant came over to me.
link |
She looked very concerned.
link |
She kind of put her hands on my shoulder.
link |
She was touching me all over, which I wasn't thrilled about
link |
and she goes, do you have some kind of anxiety disorder?
link |
And I was like, no, it's because of COVID.
link |
This is early March.
link |
Early March, but I was worried
link |
because I knew I could see or I suspected, I guess,
link |
that that doubling would continue and it did
link |
and pretty soon we had thousands of times more cases.
link |
Most of the time when I use that quote,
link |
I try to, it's motivated by more optimistic things
link |
like Moore's law and the wonders
link |
of having more computer power,
link |
but in either case, it can be very counterintuitive.
link |
I mean, if you walk for 10 minutes,
link |
you get about 10 times as far away
link |
as if you walk for one minute.
link |
That's the way our physical world works.
link |
That's the way our brains are wired,
link |
but if something doubles for 10 times as long,
link |
you don't get 10 times as much.
link |
You get a thousand times as much
link |
and after 20, it's a billion.
link |
After 30, it's a, no, sorry, after 20, it's a million.
link |
After 30, it's a billion.
link |
And pretty soon after that,
link |
it just gets to these numbers that you can barely grasp.
link |
Our world is becoming more and more exponential,
link |
mainly because of digital technologies.
link |
So more and more often our intuitions are out of whack
link |
and that can be good in the case of things creating wonders,
link |
but it can be dangerous in the case of viruses
link |
Do you think it generally applies,
link |
like is there spaces where it does apply
link |
and where it doesn't?
link |
How are we supposed to build an intuition
link |
about in which aspects of our society
link |
does exponential growth apply?
link |
Well, you can learn the math,
link |
but the truth is our brains, I think,
link |
tend to learn more from experiences.
link |
So we just start seeing it more and more often.
link |
So hanging around Silicon Valley,
link |
hanging around AI and computer researchers,
link |
I see this kind of exponential growth a lot more frequently
link |
and I'm getting used to it, but I still make mistakes.
link |
I still underestimate some of the progress
link |
in just talking to someone about GPT3
link |
and how rapidly natural language has improved.
link |
But I think that as the world becomes more exponential,
link |
we'll all start experiencing it more frequently.
link |
The danger is that we may make some mistakes in the meantime
link |
using our old kind of caveman intuitions
link |
about how the world works.
link |
Well, the weird thing is it always kind of looks linear
link |
Like it's hard to feel,
link |
it's hard to like introspect
link |
and really acknowledge how much has changed
link |
in just a couple of years or five years or 10 years
link |
with the internet.
link |
If we just look at advancements of AI
link |
or even just social media,
link |
all the various technologies
link |
that go into the digital umbrella,
link |
it feels pretty calm and normal and gradual.
link |
Well, a lot of stuff,
link |
I think there are parts of the world,
link |
most of the world that is not exponential.
link |
The way humans learn,
link |
the way organizations change,
link |
the way our whole institutions adapt and evolve,
link |
those don't improve at exponential paces.
link |
And that leads to a mismatch oftentimes
link |
between these exponentially improving technologies
link |
or let's say changing technologies
link |
because some of them are exponentially more dangerous
link |
and our intuitions and our human skills
link |
and our institutions that just don't change very fast at all.
link |
And that mismatch I think is at the root
link |
of a lot of the problems in our society,
link |
the growing inequality
link |
and other dysfunctions in our political
link |
and economic systems.
link |
So one guy that talks about exponential functions
link |
a lot is Elon Musk.
link |
He seems to internalize this kind of way
link |
of exponential thinking.
link |
He calls it first principles thinking,
link |
sort of the kind of going to the basics,
link |
asking the question,
link |
like what were the assumptions of the past?
link |
How can we throw them out the window?
link |
How can we do this 10X much more efficiently
link |
and constantly practicing that process?
link |
And also using that kind of thinking
link |
to estimate sort of when, you know, create deadlines
link |
and estimate when you'll be able to deliver
link |
on some of these technologies.
link |
Now, it often gets him in trouble
link |
because he overestimates,
link |
like he doesn't meet the initial estimates of the deadlines,
link |
but he seems to deliver late but deliver.
link |
And which is kind of interesting.
link |
Like, what are your thoughts about this whole thing?
link |
I think we can all learn from Elon.
link |
I think going to first principles,
link |
I talked about two ways of getting more of a grip
link |
on the exponential function.
link |
And one of them just comes from first principles.
link |
You know, if you understand the math of it,
link |
you can see what's gonna happen.
link |
And even if it seems counterintuitive
link |
that a couple of dozen of COVID cases
link |
can become thousands or tens or hundreds of thousands
link |
of them in a month,
link |
it makes sense once you just do the math.
link |
And I think Elon tries to do that a lot.
link |
You know, in fairness, I think he also benefits
link |
from hanging out in Silicon Valley
link |
and he's experienced it in a lot of different applications.
link |
So, you know, it's not as much of a shock to him anymore,
link |
but that's something we can all learn from.
link |
In my own life, I remember one of my first experiences
link |
really seeing it was when I was a grad student
link |
and my advisor asked me to plot the growth of computer power
link |
in the US economy in different industries.
link |
And there are all these, you know,
link |
exponentially growing curves.
link |
And I was like, holy shit, look at this.
link |
In each industry, it was just taking off.
link |
And, you know, you didn't have to be a rocket scientist
link |
to extend that and say, wow,
link |
this means that this was in the late 80s and early 90s
link |
that, you know, if it goes anything like that,
link |
we're gonna have orders of magnitude more computer power
link |
than we did at that time.
link |
And of course we do.
link |
So, you know, when people look at Moore's law,
link |
they often talk about it as just,
link |
so the exponential function is actually
link |
a stack of S curves.
link |
So basically it's you milk or whatever,
link |
take the most advantage of a particular little revolution
link |
and then you search for another revolution.
link |
And it's basically revolutions stack on top of revolutions.
link |
Do you have any intuition about how the head humans
link |
keep finding ways to revolutionize things?
link |
Well, first, let me just unpack that first point
link |
that I talked about exponential curves,
link |
but no exponential curve continues forever.
link |
It's been said that if anything can't go on forever,
link |
eventually it will stop.
link |
And, and it's very profound, but it's,
link |
it seems that a lot of people don't appreciate
link |
that half of it as well either.
link |
And that's why all exponential functions eventually turn
link |
into some kind of S curve or stop in some other way,
link |
maybe catastrophically.
link |
And that's a cap with COVID as well.
link |
I mean, it was, it went up and then it sort of, you know,
link |
at some point it starts saturating the pool of people
link |
There's a standard epidemiological model
link |
that's based on that.
link |
And it's beginning to happen with Moore's law
link |
or different generations of computer power.
link |
It happens with all exponential curves.
link |
The remarkable thing is you elude,
link |
the second part of your question is that we've been able
link |
to come up with a new S curve on top of the previous one
link |
and do that generation after generation with new materials,
link |
new processes, and just extend it further and further.
link |
I don't think anyone has a really good theory
link |
about why we've been so successful in doing that.
link |
It's great that we have been,
link |
and I hope it continues for some time,
link |
but it's, you know, one beginning of a theory
link |
is that there's huge incentives when other parts
link |
of the system are going on that clock speed
link |
of doubling every two to three years.
link |
If there's one component of it that's not keeping up,
link |
then the economic incentives become really large
link |
to improve that one part.
link |
It becomes a bottleneck and anyone who can do improvements
link |
in that part can reap huge returns
link |
so that the resources automatically get focused
link |
on whatever part of the system isn't keeping up.
link |
Do you think some version of the Moore's law will continue?
link |
Some version, yes, it is.
link |
I mean, one version that has become more important
link |
is something I call Coomey's law,
link |
which is named after John Coomey,
link |
who I should mention was also my college roommate,
link |
but he identified the fact that energy consumption
link |
has been declining by a factor of two.
link |
And for most of us, that's more important.
link |
The new iPhones came out today as we're recording this.
link |
I'm not sure when you're gonna make it available.
link |
Very soon after this, yeah.
link |
And for most of us, having the iPhone be twice as fast,
link |
it's nice, but having the battery lifelonger,
link |
that would be much more valuable.
link |
And the fact that a lot of the progress in chips now
link |
is reducing energy consumption is probably more important
link |
for many applications than just the raw speed.
link |
Other dimensions of Moore's law
link |
are in AI and machine learning.
link |
Those tend to be very parallelizable functions,
link |
especially deep neural nets.
link |
And so instead of having one chip,
link |
you can have multiple chips or you can have a GPU,
link |
graphic processing unit that goes faster.
link |
Now, special chips designed for machine learning
link |
like tensor processing units,
link |
each time you switch, there's another 10X
link |
or 100X improvement above and beyond Moore's law.
link |
So I think that the raw silicon
link |
isn't improving as much as it used to,
link |
but these other dimensions are becoming important,
link |
more important, and we're seeing progress in them.
link |
I don't know if you've seen the work by OpenAI
link |
where they show the exponential improvement
link |
of the training of neural networks
link |
just literally in the techniques used.
link |
So that's almost like the algorithm.
link |
It's fascinating to think like, can I actually continue?
link |
I was figuring out more and more tricks
link |
on how to train networks faster and faster.
link |
The progress has been staggering.
link |
If you look at image recognition, as you mentioned,
link |
I think it's a function of at least three things
link |
that are coming together.
link |
One, we just talked about faster chips,
link |
not just Moore's law, but GPUs, TPUs and other technologies.
link |
The second is just a lot more data.
link |
I mean, we are awash in digital data today
link |
in a way we weren't 20 years ago.
link |
Photography, I'm old enough to remember,
link |
it used to be chemical, and now everything is digital.
link |
I took probably 50 digital photos yesterday.
link |
I wouldn't have done that if it was chemical.
link |
And we have the internet of things
link |
and all sorts of other types of data.
link |
When we walk around with our phone,
link |
it's just broadcasting a huge amounts of digital data
link |
that can be used as training sets.
link |
And then last but not least, as they mentioned at OpenAI,
link |
there've been significant improvements in the techniques.
link |
The core idea of deep neural nets
link |
has been around for a few decades,
link |
but the advances in making it work more efficiently
link |
have also improved a couple of orders of magnitude or more.
link |
So you multiply together,
link |
a hundred fold improvement in computer power,
link |
a hundred fold or more improvement in data,
link |
a hundred fold improvement in techniques
link |
of software and algorithms,
link |
and soon you're getting into a million fold improvements.
link |
So somebody brought this up, this idea with GPT3 that,
link |
so it's trained in a self supervised way
link |
on basically internet data.
link |
And that's one of the, I've seen arguments made
link |
and they seem to be pretty convincing
link |
that the bottleneck there is going to be
link |
how much data there is on the internet,
link |
which is a fascinating idea that it literally
link |
will just run out of human generated data to train on.
link |
Right, I know we make it to the point where it's consumed
link |
basically all of human knowledge
link |
or all digitized human knowledge, yeah.
link |
And that will be the bottleneck.
link |
But the interesting thing with bottlenecks
link |
is people often use bottlenecks
link |
as a way to argue against exponential growth.
link |
They say, well, there's no way
link |
you can overcome this bottleneck,
link |
but we seem to somehow keep coming up in new ways
link |
to like overcome whatever bottlenecks
link |
the critics come up with, which is fascinating.
link |
I don't know how you overcome the data bottleneck,
link |
but probably more efficient training algorithms.
link |
Yeah, well, you already mentioned that,
link |
that these training algorithms are getting much better
link |
at using smaller amounts of data.
link |
We also are just capturing a lot more data than we used to,
link |
especially in China, but all around us.
link |
So those are both important.
link |
In some applications, you can simulate the data,
link |
video games, some of the self driving car systems
link |
are simulating driving, and of course,
link |
that has some risks and weaknesses,
link |
but you can also, if you want to exhaust
link |
all the different ways you could beat a video game,
link |
you could just simulate all the options.
link |
Can we take a step in that direction of autonomous vehicles?
link |
Next, you're talking to the CTO of Waymo tomorrow.
link |
And obviously, I'm talking to Elon again in a couple of weeks.
link |
What's your thoughts on autonomous vehicles?
link |
Like where do we stand as a problem
link |
that has the potential of revolutionizing the world?
link |
Well, I'm really excited about that,
link |
but it's become much clearer
link |
that the original way that I thought about it,
link |
most people thought about like,
link |
you know, will we have a self driving car or not
link |
is way too simple.
link |
The better way to think about it
link |
is that there's a whole continuum
link |
of how much driving and assisting the car can do.
link |
I noticed that you're right next door
link |
to the Toyota Research Institute.
link |
That is a total accident.
link |
I love the TRI folks, but yeah.
link |
Have you talked to Gil Pratt?
link |
Yeah, we're supposed to talk.
link |
It's kind of hilarious.
link |
So there's kind of the,
link |
I think it's a good counterpart to say what Elon is doing.
link |
And hopefully they can be frank
link |
in what they think about each other,
link |
because I've heard both of them talk about it.
link |
But they're much more, you know,
link |
this is an assistive, a guardian angel
link |
that watches over you as opposed to try to do everything.
link |
I think there's some things like driving on a highway,
link |
you know, from LA to Phoenix,
link |
where it's mostly good weather, straight roads.
link |
That's close to a solved problem, let's face it.
link |
In other situations, you know,
link |
driving through the snow in Boston
link |
where the roads are kind of crazy.
link |
And most importantly, you have to make a lot of judgments
link |
about what the other driver is gonna do
link |
at these intersections that aren't really right angles
link |
and aren't very well described.
link |
It's more like game theory.
link |
That's a much harder problem
link |
and requires understanding human motivations.
link |
So there's a continuum there of some places
link |
where the cars will work very well
link |
and others where it could probably take decades.
link |
What do you think about the Waymo?
link |
So you mentioned two companies
link |
that actually have cars on the road.
link |
There's the Waymo approach that it's more like
link |
we're not going to release anything until it's perfect
link |
and we're gonna be very strict
link |
about the streets that we travel on,
link |
but it better be perfect.
link |
Well, I'm smart enough to be humble
link |
and not try to get between.
link |
I know there's very bright people
link |
on both sides of the argument.
link |
I've talked to them and they make convincing arguments to me
link |
about how careful they need to be and the social acceptance.
link |
Some people thought that when the first few people died
link |
from self driving cars, that would shut down the industry,
link |
but it was more of a blip actually.
link |
And, you know, so that was interesting.
link |
Of course, there's still a concern
link |
that if there could be setbacks, if we do this wrong,
link |
you know, your listeners may be familiar
link |
with the different levels of self driving,
link |
you know, level one, two, three, four, five.
link |
I think Andrew Ng has convinced me that this idea
link |
of really focusing on level four,
link |
where you only go in areas that are well mapped
link |
rather than just going out in the wild
link |
is the way things are gonna evolve.
link |
But you can just keep expanding those areas
link |
where you've mapped things really well,
link |
where you really understand them
link |
and eventually all become kind of interconnected.
link |
And that could be a kind of another way of progressing
link |
to make it more feasible over time.
link |
I mean, that's kind of like the Waymo approach,
link |
which is they just now released,
link |
I think just like a day or two ago,
link |
a public, like anyone from the public
link |
in the Phoenix, Arizona to, you know,
link |
you can get a ride in a Waymo car
link |
with no person, no driver.
link |
Oh, they've taken away the safety driver?
link |
Oh yeah, for a while now there's been no safety driver.
link |
Okay, because I mean, I've been following that one
link |
in particular, but I thought it was kind of funny
link |
about a year ago when they had the safety driver
link |
and then they added a second safety driver
link |
because the first safety driver would fall asleep.
link |
It's like, I'm not sure they're going
link |
in the right direction with that.
link |
No, they've Waymo in particular
link |
done a really good job of that.
link |
They actually have a very interesting infrastructure
link |
of remote like observation.
link |
So they're not controlling the vehicles remotely,
link |
but they're able to, it's like a customer service.
link |
They can anytime tune into the car.
link |
I bet they can probably remotely control it as well,
link |
but that's officially not the function that they use.
link |
Yeah, I can see that being really,
link |
because I think the thing that's proven harder
link |
than maybe some of the early people expected
link |
was there's a long tail of weird exceptions.
link |
So you can deal with 90, 99, 99.99% of the cases,
link |
but then there's something that just never been seen before
link |
in the training data.
link |
And humans more or less can work around that.
link |
Although let me be clear and note,
link |
there are about 30,000 human fatalities
link |
just in the United States and maybe a million worldwide.
link |
So they're far from perfect.
link |
But I think people have higher expectations of machines.
link |
They wouldn't tolerate that level of death
link |
and damage from a machine.
link |
And so we have to do a lot better
link |
at dealing with those edge cases.
link |
And also the tricky thing that if I have a criticism
link |
for the Waymo folks, there's such a huge focus on safety
link |
where people don't talk enough about creating products
link |
that people, that customers love,
link |
that human beings love using.
link |
It's very easy to create a thing that's safe
link |
at the extremes, but then nobody wants to get into it.
link |
Yeah, well, back to Elon, I think one of,
link |
part of his genius was with the electric cars.
link |
Before he came along, electric cars were all kind of
link |
underpowered, really light,
link |
and there were sort of wimpy cars that weren't fun.
link |
And the first thing he did was he made a roadster
link |
that went zero to 60 faster than just about any other car
link |
and went the other end.
link |
And I think that was a really wise marketing move
link |
as well as a wise technology move.
link |
Yeah, it's difficult to figure out
link |
what the right marketing move is for AI systems.
link |
That's always been, I think it requires guts and risk taking
link |
which is what Elon practices.
link |
I mean, to the chagrin of perhaps investors or whatever,
link |
but it also requires rethinking what you're doing.
link |
I think way too many people are unimaginative,
link |
intellectually lazy, and when they take AI,
link |
they basically say, what are we doing now?
link |
How can we make a machine do the same thing?
link |
Maybe we'll save some costs, we'll have less labor.
link |
And yeah, it's not necessarily the worst thing
link |
in the world to do, but it's really not leading
link |
to a quantum change in the way you do things.
link |
When Jeff Bezos said, hey, we're gonna use the internet
link |
to change how bookstores work and we're gonna use technology,
link |
he didn't go and say, okay, let's put a robot cashier
link |
where the human cashier is and leave everything else alone.
link |
That would have been a very lame way to automate a bookstore.
link |
He's like went from soup to nuts and let's just rethink it.
link |
We get rid of the physical bookstore.
link |
We have a warehouse, we have delivery,
link |
we have people order on a screen
link |
and everything was reinvented.
link |
And that's been the story
link |
of these general purpose technologies all through history.
link |
And in my books, I write about like electricity
link |
and how for 30 years, there was almost no productivity gain
link |
from the electrification of factories a century ago.
link |
Now it's not because electricity
link |
is a wimpy useless technology.
link |
We all know how awesome electricity is.
link |
It's cause at first,
link |
they really didn't rethink the factories.
link |
It was only after they reinvented them
link |
and we describe how in the book,
link |
then you suddenly got a doubling and tripling
link |
of productivity growth.
link |
But it's the combination of the technology
link |
with the new business models, new business organization.
link |
That just takes a long time
link |
and it takes more creativity than most people have.
link |
Can you maybe linger on electricity?
link |
Cause that's a fun one.
link |
Yeah, well, sure, I'll tell you what happened.
link |
Before electricity, there were basically steam engines
link |
or sometimes water wheels and to power the machinery,
link |
you had to have pulleys and crankshafts
link |
and you really can't make them too long
link |
cause they'll break the torsion.
link |
So all the equipment was kind of clustered
link |
around this one giant steam engine.
link |
You can't make small steam engines either
link |
cause of thermodynamics.
link |
So you have one giant steam engine,
link |
all the equipment clustered around it, multi story.
link |
They have it vertical to minimize the distance
link |
as well as horizontal.
link |
And then when they did electricity,
link |
they took out the steam engine.
link |
They got the biggest electric motor
link |
they could buy from General Electric or someone like that.
link |
And nothing much else changed.
link |
It took until a generation of managers retired
link |
or died three years later,
link |
that people started thinking,
link |
wait, we don't have to do it that way.
link |
You can make electric motors, big, small, medium.
link |
You can put one with each piece of equipment.
link |
There's this big debate
link |
if you read the management literature
link |
between what they call a group drive versus unit drive
link |
where every machine would have its own motor.
link |
Well, once they did that, once they went to unit drive,
link |
those guys won the debate.
link |
Then you started having a new kind of factory
link |
which is sometimes spread out over acres, single story
link |
and each piece of equipment has its own motor.
link |
And most importantly, they weren't laid out based on
link |
who needed the most power.
link |
They were laid out based on
link |
what is the workflow of materials?
link |
Assembly line, let's have it go from this machine
link |
to that machine, to that machine.
link |
Once they rethought the factory that way,
link |
huge increases in productivity.
link |
It was just staggering.
link |
People like Paul David have documented this
link |
in their research papers.
link |
And I think that that is a lesson you see over and over.
link |
It happened when the steam engine changed manual production.
link |
It's happened with the computerization.
link |
People like Michael Hammer said, don't automate, obliterate.
link |
In each case, the big gains only came once
link |
smart entrepreneurs and managers
link |
basically reinvented their industries.
link |
I mean, one other interesting point about all that
link |
is that during that reinvention period,
link |
you often actually not only don't see productivity growth,
link |
you can actually see a slipping back.
link |
Measured productivity actually falls.
link |
I just wrote a paper with Chad Severson and Daniel Rock
link |
called the productivity J curve,
link |
which basically shows that in a lot of these cases,
link |
you have a downward dip before it goes up.
link |
And that downward dip is when everyone's trying
link |
to like reinvent things.
link |
And you could say that they're creating knowledge
link |
and intangible assets,
link |
but that doesn't show up on anyone's balance sheet.
link |
It doesn't show up in GDP.
link |
So it's as if they're doing nothing.
link |
Like take self driving cars, we were just talking about it.
link |
There have been hundreds of billions of dollars
link |
spent developing self driving cars.
link |
And basically no chauffeur has lost his job, no taxi driver.
link |
I guess I got to check out the ones that.
link |
It's a big J curve.
link |
Yeah, so there's a bunch of spending
link |
and no real consumer benefit.
link |
Now they're doing that in the belief,
link |
I think the justified belief
link |
that they will get the upward part of the J curve
link |
and there will be some big returns,
link |
but in the short run, you're not seeing it.
link |
That's happening with a lot of other AI technologies,
link |
just as it happened
link |
with earlier general purpose technologies.
link |
And it's one of the reasons
link |
we're having relatively low productivity growth lately.
link |
As an economist, one of the things that disappoints me
link |
is that as eye popping as these technologies are,
link |
you and I are both excited
link |
about some of the things they can do.
link |
The economic productivity statistics are kind of dismal.
link |
We actually, believe it or not,
link |
have had lower productivity growth
link |
in the past about 15 years
link |
than we did in the previous 15 years,
link |
in the 90s and early 2000s.
link |
And so that's not what you would have expected
link |
if these technologies were that much better.
link |
But I think we're in kind of a long J curve there.
link |
Personally, I'm optimistic.
link |
We'll start seeing the upward tick,
link |
maybe as soon as next year.
link |
But the past decade has been a bit disappointing
link |
if you thought there's a one to one relationship
link |
between cool technology and higher productivity.
link |
Well, what would you place your biggest hope
link |
for productivity increases on?
link |
Because you kind of said at a high level AI,
link |
but if I were to think about
link |
what has been so revolutionary in the last 10 years,
link |
I would 15 years and thinking about the internet,
link |
I would say things like,
link |
hopefully I'm not saying anything ridiculous,
link |
but everything from Wikipedia to Twitter.
link |
So like these kind of websites,
link |
but like I would expect to see some kind
link |
of big productivity increases
link |
from just the connectivity between people
link |
and the access to more information.
link |
Yeah, well, so that's another area
link |
I've done quite a bit of research on actually,
link |
is these free goods like Wikipedia, Facebook, Twitter, Zoom.
link |
We're actually doing this in person,
link |
but almost everything else I do these days is online.
link |
The interesting thing about all those
link |
is most of them have a price of zero.
link |
What do you pay for Wikipedia?
link |
Maybe like a little bit for the electrons
link |
to come to your house.
link |
Basically zero, right?
link |
Take a small pause and say, I donate to Wikipedia.
link |
Often you should too.
link |
It's good for you, yeah.
link |
So, but what does that do mean for GDP?
link |
GDP is based on the price and quantity
link |
of all the goods, things bought and sold.
link |
If something has zero price,
link |
you know how much it contributes to GDP?
link |
To a first approximation, zero.
link |
So these digital goods that we're getting more and more of,
link |
we're spending more and more hours a day
link |
consuming stuff off of screens,
link |
little screens, big screens,
link |
that doesn't get priced into GDP.
link |
It's like they don't exist.
link |
That doesn't mean they don't create value.
link |
I get a lot of value from watching cat videos
link |
and reading Wikipedia articles and listening to podcasts,
link |
even if I don't pay for them.
link |
So we've got a mismatch there.
link |
Now, in fairness, economists,
link |
since Simon Kuznets invented GDP and productivity,
link |
all those statistics back in the 1930s,
link |
he recognized, he in fact said,
link |
this is not a measure of wellbeing.
link |
This is not a measure of welfare.
link |
It's a measure of production.
link |
But almost everybody has kind of forgotten
link |
that he said that and they just use it.
link |
It's like, how well off are we?
link |
What was GDP last year?
link |
It was 2.3% growth or whatever.
link |
That is how much physical production,
link |
but it's not the value we're getting.
link |
We need a new set of statistics
link |
and I'm working with some colleagues.
link |
Avi Collis and others to develop something
link |
we call GDP dash B.
link |
GDP B measures the benefits you get, not the cost.
link |
If you get benefit from Zoom or Wikipedia or Facebook,
link |
then that gets counted in GDP B,
link |
even if you pay zero for it.
link |
So, you know, back to your original point,
link |
I think there is a lot of gain over the past decade
link |
in these digital goods that doesn't show up in GDP,
link |
doesn't show up in productivity.
link |
By the way, productivity is just defined
link |
as GDP divided by hours worked.
link |
So if you mismeasure GDP,
link |
you mismeasure productivity by the exact same amount.
link |
That's something we need to fix.
link |
I'm working with the statistical agencies
link |
to come up with a new set of metrics.
link |
And, you know, over the coming years,
link |
I think we'll see, we're not gonna do away with GDP.
link |
It's very useful, but we'll see a parallel set of accounts
link |
that measure the benefits.
link |
How difficult is it to get that B in the GDP B?
link |
I mean, one of the reasons it hasn't been done before
link |
is that, you know, you can measure it,
link |
the cash register, what people pay for stuff,
link |
but how do you measure what they would have paid,
link |
like what the value is?
link |
That's a lot harder, you know?
link |
How much is Wikipedia worth to you?
link |
That's what we have to answer.
link |
And to do that, what we do is we can use online experiments.
link |
We do massive online choice experiments.
link |
We ask hundreds of thousands, now millions of people
link |
to do lots of sort of A, B tests.
link |
How much would I have to pay you
link |
to give up Wikipedia for a month?
link |
How much would I have to pay you to stop using your phone?
link |
And in some cases, it's hypothetical.
link |
In other cases, we actually enforce it,
link |
which is kind of expensive.
link |
Like we pay somebody $30 to stop using Facebook
link |
and we see if they'll do it.
link |
And some people will give it up for $10.
link |
Some people won't give it up even if you give them $100.
link |
And then you get a whole demand curve.
link |
You get to see what all the different prices are
link |
and how much value different people get.
link |
And not surprisingly,
link |
different people have different values.
link |
We find that women tend to value Facebook more than men.
link |
Old people tend to value it a little bit more
link |
than young people.
link |
That was interesting.
link |
I think young people maybe know about other networks
link |
that I don't know the name of that are better than Facebook.
link |
And so you get to see these patterns,
link |
but every person's individual.
link |
And then if you add up all those numbers,
link |
you start getting an estimate of the value.
link |
Okay, first of all, that's brilliant.
link |
Is this a work that will soon eventually be published?
link |
Yeah, well, there's a version of it
link |
in the Proceedings of the National Academy of Sciences
link |
about I think we call it massive online choice experiments.
link |
I should remember the title, but it's on my website.
link |
So yeah, we have some more papers coming out on it,
link |
but the first one is already out.
link |
You know, it's kind of a fascinating mystery
link |
that Twitter, Facebook,
link |
like all these social networks are free.
link |
And it seems like almost none of them except for YouTube
link |
have experimented with removing ads for money.
link |
Can you like, do you understand that
link |
from both economics and the product perspective?
link |
Yeah, it's something that, you know,
link |
so I teach a course on digital business models.
link |
So I used to at MIT, at Stanford, I'm not quite sure.
link |
I'm not teaching until next spring.
link |
I'm still thinking what my course is gonna be.
link |
But there are a lot of different business models.
link |
And when you have something that has zero marginal cost,
link |
there's a lot of forces,
link |
especially if there's any kind of competition
link |
that push prices down to zero.
link |
But you can have ad supported systems,
link |
you can bundle things together.
link |
You can have volunteer, you mentioned Wikipedia,
link |
there's donations.
link |
And I think economists underestimate
link |
the power of volunteerism and donations.
link |
Your national public radio.
link |
Actually, how do you, this podcast, how is this,
link |
what's the revenue model?
link |
There's sponsors at the beginning.
link |
And then, and people, the funny thing is,
link |
I tell people they can, it's very,
link |
I tell them the timestamp.
link |
So if you wanna skip the sponsors, you're free.
link |
But it's funny that a bunch of people,
link |
so I read the advertisement
link |
and then a bunch of people enjoy reading it.
link |
Well, they may learn something from it.
link |
And also from the advertiser's perspective,
link |
those are people who are actually interested.
link |
I mean, the example I sometimes get is like,
link |
I bought a car recently and all of a sudden,
link |
all the car ads were like interesting to me.
link |
And then like, now that I have the car,
link |
like I sort of zone out on, but that's fine.
link |
The car companies, they don't really wanna be advertising
link |
to me if I'm not gonna buy their product.
link |
So there are a lot of these different revenue models
link |
and it's a little complicated,
link |
but the economic theory has to do
link |
with what the shape of the demand curve is,
link |
when it's better to monetize it with charging people
link |
versus when you're better off doing advertising.
link |
I mean, in short, when the demand curve
link |
is relatively flat and wide,
link |
like generic news and things like that,
link |
then you tend to do better with advertising.
link |
If it's a good that's only useful to a small number
link |
of people, but they're willing to pay a lot,
link |
they have a very high value for it,
link |
then advertising isn't gonna work as well
link |
and you're better off charging for it.
link |
Both of them have some inefficiencies.
link |
And then when you get into targeting
link |
and you get into these other revenue models,
link |
it gets more complicated,
link |
but there's some economic theory on it.
link |
I also think to be frank,
link |
there's just a lot of experimentation that's needed
link |
because sometimes things are a little counterintuitive,
link |
especially when you get into what are called
link |
two sided networks or platform effects,
link |
where you may grow the market on one side
link |
and harvest the revenue on the other side.
link |
Facebook tries to get more and more users
link |
and then they harvest the revenue from advertising.
link |
So that's another way of kind of thinking about it.
link |
Is it strange to you that they haven't experimented?
link |
Well, they are experimenting.
link |
So they are doing some experiments
link |
about what the willingness is for people to pay.
link |
I think that when they do the math,
link |
it's gonna work out that they still are better off
link |
with an advertising driven model, but...
link |
Like this is what YouTube is, right?
link |
It's you allow the person to decide,
link |
the customer to decide exactly which model they prefer.
link |
No, that can work really well.
link |
And newspapers, of course,
link |
have known this for a long time.
link |
The Wall Street Journal, the New York Times,
link |
they have subscription revenue.
link |
They also have advertising revenue.
link |
And that can definitely work.
link |
Online, it's a lot easier to have a dial
link |
that's much more personalized
link |
and everybody can kind of roll their own mix.
link |
And I could imagine having a little slider
link |
about how much advertising you want or are willing to take.
link |
And if it's done right and it's incentive compatible,
link |
it could be a win win where both the content provider
link |
and the consumer are better off
link |
than they would have been before.
link |
Yeah, the done right part is a really good point.
link |
Like with the Jeff Bezos
link |
and the single click purchase on Amazon,
link |
the frictionless effort there,
link |
if I could just rant for a second
link |
about the Wall Street Journal,
link |
all the newspapers you mentioned,
link |
is I have to click so many times to subscribe to them
link |
that I literally don't subscribe
link |
just because of the number of times I have to click.
link |
I'm totally with you.
link |
I don't understand why so many companies make it so hard.
link |
I mean, another example is when you buy a new iPhone
link |
or a new computer, whatever,
link |
I feel like, okay, I'm gonna lose an afternoon
link |
just like loading up and getting all my stuff back.
link |
And for a lot of us,
link |
that's more of a deterrent than the price.
link |
And if they could make it painless,
link |
we'd give them a lot more money.
link |
So I'm hoping somebody listening is working
link |
on making it more painless for us to buy your products.
link |
If we could just like linger a little bit
link |
on the social network thing,
link |
because there's this Netflix social dilemma.
link |
Yeah, no, I saw that.
link |
And Tristan Harris and company, yeah.
link |
And people's data,
link |
it's really sensitive and social networks
link |
are at the core arguably of many of societal like tension
link |
and some of the most important things happening in society.
link |
So it feels like it's important to get this right,
link |
both from a business model perspective
link |
and just like a trust perspective.
link |
I still gotta, I mean, it just still feels like,
link |
I know there's experimentation going on.
link |
It still feels like everyone is afraid
link |
to try different business models, like really try.
link |
Well, I'm worried that people are afraid
link |
to try different business models.
link |
I'm also worried that some of the business models
link |
may lead them to bad choices.
link |
And Danny Kahneman talks about system one and system two,
link |
sort of like a reptilian brain
link |
that reacts quickly to what we see,
link |
see something interesting, we click on it,
link |
we retweet it versus our system two,
link |
our frontal cortex that's supposed to be more careful
link |
and rational that really doesn't make
link |
as many decisions as it should.
link |
I think there's a tendency for a lot of these social networks
link |
to really exploit system one, our quick instant reaction,
link |
make it so we just click on stuff and pass it on
link |
and not really think carefully about it.
link |
And that system, it tends to be driven
link |
by sex, violence, disgust, anger, fear,
link |
these relatively primitive kinds of emotions.
link |
Maybe they're important for a lot of purposes,
link |
but they're not a great way to organize a society.
link |
And most importantly, when you think about this huge,
link |
amazing information infrastructure we've had
link |
that's connected billions of brains across the globe,
link |
not just so we can all access information,
link |
but we can all contribute to it and share it.
link |
Arguably the most important thing
link |
that that network should do is favor truth over falsehoods.
link |
And the way it's been designed,
link |
not necessarily intentionally, is exactly the opposite.
link |
My MIT colleagues are all, and Deb Roy and others at MIT,
link |
did a terrific paper in the cover of Science.
link |
And they documented what we all feared,
link |
which is that lies spread faster than truth
link |
on social networks.
link |
They looked at a bunch of tweets and retweets,
link |
and they found that false information
link |
was more likely to spread further, faster, to more people.
link |
It's not because people like lies.
link |
It's because people like things that are shocking,
link |
amazing, can you believe this?
link |
Something that is not mundane,
link |
not something that everybody else already knew.
link |
And what are the most unbelievable things?
link |
And so if you wanna find something unbelievable,
link |
it's a lot easier to do that
link |
if you're not constrained by the truth.
link |
So they found that the emotional valence
link |
of false information was just much higher.
link |
It was more likely to be shocking,
link |
and therefore more likely to be spread.
link |
Another interesting thing was that
link |
that wasn't necessarily driven by the algorithms.
link |
I know that there is some evidence,
link |
Zeynep Tufekci and others have pointed out on YouTube,
link |
some of the algorithms unintentionally were tuned
link |
to amplify more extremist content.
link |
But in the study of Twitter that Sinan and Deb and others did,
link |
they found that even if you took out all the bots
link |
and all the automated tweets,
link |
you still had lies spreading significantly faster.
link |
It's just the problems with ourselves
link |
that we just can't resist passing on the salacious content.
link |
But I also blame the platforms
link |
because there's different ways you can design a platform.
link |
You can design a platform in a way
link |
that makes it easy to spread lies
link |
and to retweet and spread things on,
link |
or you can kind of put some friction on that
link |
and try to favor truth.
link |
I had dinner with Jimmy Wales once,
link |
the guy who helped found Wikipedia.
link |
And he convinced me that, look,
link |
you can make some design choices,
link |
whether it's at Facebook, at Twitter,
link |
at Wikipedia, or Reddit, whatever,
link |
and depending on how you make those choices,
link |
you're more likely or less likely to have false news.
link |
Create a little bit of friction, like you said.
link |
You know, that's the, and so if I'm...
link |
It could be friction, it could be speeding the truth,
link |
either way, but, and I don't totally understand...
link |
Speeding the truth, I love it.
link |
Amplifying it and giving it more credit.
link |
And in academia, which is far, far from perfect,
link |
but when someone has an important discovery,
link |
it tends to get more cited
link |
and people kind of look to it more
link |
and sort of, it tends to get amplified a little bit.
link |
So you could try to do that too.
link |
I don't know what the silver bullet is,
link |
but the meta point is that if we spend time
link |
thinking about it, we can amplify truth over falsehoods.
link |
And I'm disappointed in the heads of these social networks
link |
that they haven't been as successful
link |
or maybe haven't tried as hard to amplify truth.
link |
And part of it, going back to what we said earlier,
link |
is these revenue models may push them
link |
more towards growing fast, spreading information rapidly,
link |
getting lots of users,
link |
which isn't the same thing as finding truth.
link |
Yeah, I mean, implicit in what you're saying now
link |
is a hopeful message that with platforms,
link |
we can take a step towards a greater
link |
and greater popularity of truth.
link |
But the more cynical view is that
link |
what the last few years have revealed
link |
is that there's a lot of money to be made
link |
in dismantling even the idea of truth,
link |
that nothing is true.
link |
And as a thought experiment,
link |
I've been thinking about if it's possible
link |
that our future will have,
link |
like the idea of truth is something we won't even have.
link |
Do you think it's possible in the future
link |
that everything is on the table in terms of truth,
link |
and we're just swimming in this kind of digital economy
link |
where ideas are just little toys
link |
that are not at all connected to reality?
link |
Yeah, I think that's definitely possible.
link |
I'm not a technological determinist,
link |
so I don't think that's inevitable.
link |
I don't think it's inevitable that it doesn't happen.
link |
I mean, the thing that I've come away with
link |
every time I do these studies,
link |
and I emphasize it in my books and elsewhere,
link |
is that technology doesn't shape our destiny,
link |
we shape our destiny.
link |
So just by us having this conversation,
link |
I hope that your audience is gonna take it upon themselves
link |
as they design their products,
link |
and they think about, they use products,
link |
as they manage companies,
link |
how can they make conscious decisions
link |
to favor truth over falsehoods,
link |
favor the better kinds of societies,
link |
and not abdicate and say, well, we just build the tools.
link |
I think there was a saying that,
link |
was it the German scientist
link |
when they were working on the missiles in late World War II?
link |
They said, well, our job is to make the missiles go up.
link |
Where they come down, that's someone else's department.
link |
And that's obviously not the, I think it's obvious,
link |
that's not the right attitude
link |
that technologists should have,
link |
that engineers should have.
link |
They should be very conscious
link |
about what the implications are.
link |
And if we think carefully about it,
link |
we can avoid the kind of world that you just described,
link |
where truth is all relative.
link |
There are going to be people who benefit from a world
link |
of where people don't check facts,
link |
and where truth is relative,
link |
and popularity or fame or money is orthogonal to truth.
link |
But one of the reasons I suspect
link |
that we've had so much progress over the past few hundred
link |
years is the invention of the scientific method,
link |
which is a really powerful tool or meta tool
link |
for finding truth and favoring things that are true
link |
versus things that are false.
link |
If they don't pass the scientific method,
link |
they're less likely to be true.
link |
And that has, the societies and the people
link |
and the organizations that embrace that
link |
have done a lot better than the ones who haven't.
link |
And so I'm hoping that people keep that in mind
link |
and continue to try to embrace not just the truth,
link |
but methods that lead to the truth.
link |
So maybe on a more personal question,
link |
if one were to try to build a competitor to Twitter,
link |
what would you advise?
link |
Is there, I mean, the bigger, the meta question,
link |
is that the right way to improve systems?
link |
Yeah, no, I think that the underlying premise
link |
behind Twitter and all these networks is amazing,
link |
that we can communicate with each other.
link |
And I use it a lot.
link |
There's a subpart of Twitter called Econ Twitter,
link |
where we economists tweet to each other
link |
and talk about new papers.
link |
Something came out in the NBER,
link |
the National Bureau of Economic Research,
link |
and we share about it.
link |
People critique it.
link |
I think it's been a godsend
link |
because it's really sped up the scientific process,
link |
if you can call economic scientific.
link |
Does it get divisive in that little?
link |
Sometimes, yeah, sure.
link |
Sometimes it does.
link |
It can also be done in nasty ways and there's the bad parts.
link |
But the good parts are great
link |
because you just speed up that clock speed
link |
of learning about things.
link |
Instead of like in the old, old days,
link |
waiting to read it in a journal,
link |
or the not so old days when you'd see it posted
link |
on a website and you'd read it.
link |
Now on Twitter, people will distill it down
link |
and it's a real art to getting to the essence of things.
link |
So that's been great.
link |
But it certainly, we all know that Twitter
link |
can be a cesspool of misinformation.
link |
And like I just said,
link |
unfortunately misinformation tends to spread faster
link |
on Twitter than truth.
link |
And there are a lot of people
link |
who are very vulnerable to it.
link |
I'm sure I've been fooled at times.
link |
There are agents, whether from Russia
link |
or from political groups or others
link |
that explicitly create efforts at misinformation
link |
and efforts at getting people to hate each other.
link |
Or even more important lately I've discovered
link |
You know the idea of nut picking?
link |
Nut picking is when you find like an extreme nut case
link |
on the other side and then you amplify them
link |
and make it seem like that's typical of the other side.
link |
So you're not literally lying.
link |
You're taking some idiot, you know,
link |
renting on the subway or just, you know,
link |
whether they're in the KKK or Antifa or whatever,
link |
they're just, and you,
link |
normally nobody would pay attention to this guy.
link |
Like 12 people would see him and it'd be the end.
link |
Instead with video or whatever,
link |
you get tens of millions of people say it.
link |
And I've seen this, you know, I look at it,
link |
I'm like, I get angry.
link |
I'm like, I can't believe that person
link |
did something so terrible.
link |
Let me tell all my friends about this terrible person.
link |
And it's a great way to generate division.
link |
I talked to a friend who studied Russian misinformation
link |
campaigns, and they're very clever about literally
link |
being on both sides of some of these debates.
link |
They would have some people pretend to be part of BLM.
link |
Some people pretend to be white nationalists
link |
and they would be throwing epithets at each other,
link |
saying crazy things at each other.
link |
And they're literally playing both sides of it,
link |
but their goal wasn't for one or the other to win.
link |
It was for everybody to get behaving
link |
and distrusting everyone else.
link |
So these tools can definitely be used for that.
link |
And they are being used for that.
link |
It's been super destructive for our democracy
link |
And the people who run these platforms,
link |
I think have a social responsibility,
link |
a moral and ethical, personal responsibility
link |
to do a better job and to shut that stuff down.
link |
Well, I don't know if you can shut it down,
link |
but to design them in a way that, you know,
link |
as I said earlier, favors truth over falsehoods
link |
and favors positive types of
link |
communication versus destructive ones.
link |
And just like you said, it's also on us.
link |
I try to be all about love and compassion,
link |
empathy on Twitter.
link |
I mean, one of the things,
link |
nut picking is a fascinating term.
link |
One of the things that people do,
link |
that's I think even more dangerous
link |
is nut picking applied to individual statements
link |
So basically worst case analysis in computer science
link |
is taking sometimes out of context,
link |
but sometimes in context,
link |
a statement, one statement by a person,
link |
like I've been, because I've been reading
link |
The Rise and Fall of the Third Reich,
link |
I often talk about Hitler on this podcast with folks
link |
and it is so easy.
link |
That's really dangerous.
link |
But I'm all leaning in, I'm 100%.
link |
Because, well, it's actually a safer place
link |
than people realize because it's history
link |
and history in long form is actually very fascinating
link |
to think about and it's,
link |
but I could see how that could be taken
link |
totally out of context and it's very worrying.
link |
You know, these digital infrastructures,
link |
not just they disseminate things,
link |
but they're sort of permanent.
link |
So anything you say at some point,
link |
someone can go back and find something you said
link |
three years ago, perhaps jokingly, perhaps not,
link |
maybe you're just wrong and you made them, you know,
link |
and like that becomes, they can use that to define you
link |
if they have ill intent.
link |
And we all need to be a little more forgiving.
link |
I mean, somewhere in my 20s, I told myself,
link |
I was going through all my different friends
link |
and I was like, you know, every one of them
link |
has at least like one nutty opinion.
link |
And I was like, there's like nobody
link |
who's like completely, except me, of course,
link |
but I'm sure they thought that about me too.
link |
And so you just kind of like learned
link |
to be a little bit tolerant that like, okay,
link |
there's just, you know.
link |
Yeah, I wonder who the responsibility lays on there.
link |
Like, I think ultimately it's about leadership.
link |
Like the previous president, Barack Obama,
link |
has been, I think, quite eloquent
link |
at walking this very difficult line
link |
of talking about cancel culture, but it's a difficult,
link |
Because you say the wrong thing
link |
and you piss off a lot of people.
link |
And so you have to do it well.
link |
But then also the platform of the technology is,
link |
should slow down, create friction,
link |
and spreading this kind of nut picking in all its forms.
link |
No, and your point that we have to like learn over time,
link |
I mean, we can't put it all on the platform
link |
and say, you guys design it.
link |
Because if we're idiots about using it,
link |
nobody can design a platform that withstands that.
link |
And every new technology people learn its dangers.
link |
You know, when someone invented fire,
link |
it's great cooking and everything,
link |
but then somebody burned themself.
link |
And then you had to like learn how to like avoid,
link |
maybe somebody invented a fire extinguisher later.
link |
So you kind of like figure out ways
link |
of working around these technologies.
link |
Someone invented seat belts, et cetera.
link |
And that's certainly true
link |
with all the new digital technologies
link |
that we have to figure out,
link |
not just technologies that protect us,
link |
but ways of using them that emphasize
link |
that are more likely to be successful than dangerous.
link |
So you've written quite a bit
link |
about how artificial intelligence might change our world.
link |
How do you think if we look forward,
link |
again, it's impossible to predict the future,
link |
but if we look at trends from the past
link |
and we tried to predict what's gonna happen
link |
in the rest of the 21st century,
link |
how do you think AI will change our world?
link |
That's a big question.
link |
You know, I'm mostly a techno optimist.
link |
I'm not at the extreme, you know,
link |
the singularity is near end of the spectrum,
link |
but I do think that we're likely in
link |
for some significantly improved living standards,
link |
some really important progress,
link |
even just the technologies that are already kind of like
link |
in the can that haven't diffused.
link |
You know, when I talked earlier about the J curve,
link |
it could take 10, 20, 30 years for an existing technology
link |
to have the kind of profound effects.
link |
And when I look at whether it's, you know,
link |
vision systems, voice recognition, problem solving systems,
link |
even if nothing new got invented,
link |
we would have a few decades of progress.
link |
So I'm excited about that.
link |
And I think that's gonna lead to us being wealthier,
link |
healthier, I mean,
link |
the healthcare is probably one of the applications
link |
that I'm most excited about.
link |
So that's good news.
link |
I don't think we're gonna have the end of work anytime soon.
link |
There's just too many things that machines still can't do.
link |
When I look around the world
link |
and think of whether it's childcare or healthcare,
link |
cleaning the environment, interacting with people,
link |
scientific work, artistic creativity,
link |
these are things that for now,
link |
machines aren't able to do nearly as well as humans,
link |
even just something as mundane as, you know,
link |
folding laundry or whatever.
link |
And many of these, I think are gonna be years or decades
link |
before machines catch up.
link |
You know, I may be surprised on some of them,
link |
but overall, I think there's plenty of work
link |
There's plenty of problems in society
link |
that need the human touch.
link |
So we'll have to repurpose.
link |
We'll have to, as machines are able to do some tasks,
link |
people are gonna have to reskill and move into other areas.
link |
And that's probably what's gonna be going on
link |
for the next, you know, 10, 20, 30 years or more,
link |
kind of big restructuring of society.
link |
We'll get wealthier and people will have to do new skills.
link |
Now, if you turn the dial further, I don't know,
link |
50 or a hundred years into the future,
link |
then, you know, maybe all bets are off.
link |
Then it's possible that machines will be able to do
link |
most of what people do.
link |
You know, say one or 200 years, I think it's even likely.
link |
And at that point,
link |
then we're more in the sort of abundance economy.
link |
Then we're in a world where there's really little
link |
for the humans can do economically better than machines,
link |
other than be human.
link |
And, you know, that will take a transition as well,
link |
kind of more of a transition of how we get meaning in life
link |
and what our values are.
link |
But shame on us if we screw that up.
link |
I mean, that should be like great, great news.
link |
And it kind of saddens me that some people see that
link |
as like a big problem.
link |
I think that would be, should be wonderful
link |
if people have all the health and material things
link |
that they need and can focus on loving each other
link |
and discussing philosophy and playing
link |
and doing all the other things that don't require work.
link |
Do you think you'd be surprised to see what the 20,
link |
if we were to travel in time, 100 years into the future,
link |
do you think you'll be able to,
link |
like if I gave you a month to like talk to people,
link |
no, like let's say a week,
link |
would you be able to understand what the hell's going on?
link |
You mean if I was there for a week?
link |
Yeah, if you were there for a week.
link |
A hundred years in the future?
link |
So like, so I'll give you one thought experiment is like,
link |
isn't it possible that we're all living in virtual reality
link |
Yeah, no, I think that's very possible.
link |
I've played around with some of those VR headsets
link |
and they're not great,
link |
but I mean the average person spends many waking hours
link |
staring at screens right now.
link |
They're kind of low res compared to what they could be
link |
in 30 or 50 years, but certainly games
link |
and why not any other interactions could be done with VR?
link |
And that would be a pretty different world
link |
and we'd all, in some ways be as rich as we wanted.
link |
We could have castles and we could be traveling
link |
anywhere we want and it could obviously be multisensory.
link |
So that would be possible and of course there's people,
link |
you've had Elon Musk on and others, there are people,
link |
Nick Bostrom makes the simulation argument
link |
that maybe we're already there.
link |
We're already there.
link |
So, but in general, or do you not even think about
link |
in this kind of way, you're self critically thinking,
link |
how good are you as an economist at predicting
link |
what the future looks like?
link |
Well, it starts getting, I mean,
link |
I feel reasonably comfortable the next five, 10, 20 years
link |
in terms of that path.
link |
When you start getting truly superhuman
link |
artificial intelligence, kind of by definition,
link |
be able to think of a lot of things
link |
that I couldn't have thought of and create a world
link |
that I couldn't even imagine.
link |
And so I'm not sure I can predict what that world
link |
is going to be like.
link |
One thing that AI researchers, AI safety researchers
link |
worry about is what's called the alignment problem.
link |
When an AI is that powerful,
link |
then they can do all sorts of things.
link |
And you really hope that their values
link |
are aligned with our values.
link |
And it's even tricky to finding what our values are.
link |
I mean, first off, we all have different values.
link |
And secondly, maybe if we were smarter,
link |
we would have better values.
link |
Like, I like to think that we have better values
link |
than we did in 1860 and, or in the year 200 BC
link |
on a lot of dimensions,
link |
things that we consider barbaric today.
link |
And it may be that if I thought about it more deeply,
link |
I would also be morally evolved.
link |
Maybe I'd be a vegetarian or do other things
link |
that right now, whether my future self
link |
would consider kind of immoral.
link |
So that's a tricky problem,
link |
getting the AI to do what we want,
link |
assuming it's even a friendly AI.
link |
I mean, I should probably mention
link |
there's a nontrivial other branch
link |
where we destroy ourselves, right?
link |
I mean, there's a lot of exponentially improving
link |
technologies that could be ferociously destructive,
link |
whether it's in nanotechnology or biotech
link |
and weaponized viruses, AI and other things that.
link |
Nuclear weapons, of course.
link |
The old school technology.
link |
Yeah, good old nuclear weapons that could be devastating
link |
or even existential and new things yet to be invented.
link |
So that's a branch that I think is pretty significant.
link |
And there are those who think that one of the reasons
link |
we haven't been contacted by other civilizations, right?
link |
Is that once you get to a certain level of complexity
link |
in technology, there's just too many ways to go wrong.
link |
There's a lot of ways to blow yourself up.
link |
And people, or I should say species,
link |
end up falling into one of those traps.
link |
I mean, there's an optimistic view of that.
link |
If there is literally no intelligent life out there
link |
in the universe, or at least in our galaxy,
link |
that means that we've passed at least one
link |
of the great filters or some of the great filters
link |
Yeah, no, I think Robin Hansen has a good way of,
link |
maybe others have a good way of thinking about this,
link |
that if there are no other intelligence creatures out there
link |
that we've been able to detect,
link |
one possibility is that there's a filter ahead of us.
link |
And when you get a little more advanced,
link |
maybe in a hundred or a thousand or 10,000 years,
link |
things just get destroyed for some reason.
link |
The other one is the great filters behind us.
link |
That'll be good, is that most planets don't even evolve life
link |
or if they don't evolve life,
link |
they don't evolve intelligent life.
link |
Maybe we've gotten past that.
link |
And so now maybe we're on the good side
link |
of the great filter.
link |
So if we sort of rewind back and look at the thing
link |
where we could say something a little bit more comfortably
link |
at five years and 10 years out,
link |
you've written about jobs
link |
and the impact on sort of our economy and the jobs
link |
in terms of artificial intelligence that it might have.
link |
It's a fascinating question of what kind of jobs are safe,
link |
what kind of jobs are not.
link |
Can you maybe speak to your intuition
link |
about how we should think about AI changing
link |
the landscape of work?
link |
Well, this is a really important question
link |
because I think we're very far
link |
from artificial general intelligence,
link |
which is AI that can just do the full breadth
link |
of what humans can do.
link |
But we do have human level or superhuman level
link |
narrow intelligence, narrow artificial intelligence.
link |
And obviously my calculator can do math a lot better
link |
And there's a lot of other things
link |
that machines can do better than I can.
link |
So which is which?
link |
We actually set out to address that question
link |
with Tom Mitchell.
link |
I wrote a paper called what can machine learning do
link |
that was in science.
link |
And we went and interviewed a whole bunch of AI experts
link |
and kind of synthesized what they thought machine learning
link |
was good at and wasn't good at.
link |
And we came up with what we called a rubric,
link |
basically a set of questions you can ask about any task
link |
that will tell you whether it's likely to score high or low
link |
on suitability for machine learning.
link |
And then we've applied that
link |
to a bunch of tasks in the economy.
link |
In fact, there's a data set of all the tasks
link |
in the US economy, believe it or not, it's called ONET.
link |
The US government put it together,
link |
part of the Bureau of Labor Statistics.
link |
They divide the economy into about 970 occupations
link |
like bus driver, economist, primary school teacher,
link |
radiologist, and then for each one of them,
link |
they describe which tasks need to be done.
link |
Like for radiologists, there are 27 distinct tasks.
link |
So we went through all those tasks
link |
to see whether or not a machine could do them.
link |
And what we found interestingly was...
link |
Brilliant study by the way, that's so awesome.
link |
So what we found was that there was no occupation
link |
in our data set where machine learning just ran the table
link |
and did everything.
link |
And there was almost no occupation
link |
where machine learning didn't have
link |
like a significant ability to do things.
link |
Like take radiology, a lot of people I hear saying,
link |
you know, it's the end of radiology.
link |
And one of the 27 tasks is read medical images.
link |
Really important one, like it's kind of a core job.
link |
And machines have basically gotten as good
link |
or better than radiologists.
link |
There was just an article in Nature last week,
link |
but they've been publishing them for the past few years
link |
showing that machine learning can do as well as humans
link |
on many kinds of diagnostic imaging tasks.
link |
But other things that radiologists do,
link |
they sometimes administer conscious sedation.
link |
They sometimes do physical exams.
link |
They have to synthesize the results
link |
and explain it to the other doctors or to the patients.
link |
In all those categories,
link |
machine learning isn't really up to snuff yet.
link |
So that job, we're gonna see a lot of restructuring.
link |
Parts of the job, they'll hand over to machines.
link |
Others, humans will do more of.
link |
That's been more or less the pattern all of them.
link |
So, you know, to oversimplify a bit,
link |
we're gonna see a lot of restructuring,
link |
reorganization of work.
link |
And it's real gonna be a great time.
link |
It is a great time for smart entrepreneurs and managers
link |
to do that reinvention of work.
link |
I'm not gonna see mass unemployment.
link |
To get more specifically to your question,
link |
the kinds of tasks that machines tend to be good at
link |
are a lot of routine problem solving,
link |
mapping inputs X into outputs Y.
link |
If you have a lot of data on the Xs and the Ys,
link |
the inputs and the outputs,
link |
you can do that kind of mapping and find the relationships.
link |
They tend to not be very good at,
link |
even now, fine motor control and dexterity.
link |
Emotional intelligence and human interactions
link |
and thinking outside the box, creative work.
link |
If you give it a well structured task,
link |
machines can be very good at it.
link |
But even asking the right questions, that's hard.
link |
There's a quote that Andrew McAfee and I use
link |
in our book, Second Machine Age.
link |
Apparently Pablo Picasso was shown an early computer
link |
and he came away kind of unimpressed.
link |
He goes, well, I don't see all the fusses.
link |
All that does is answer questions.
link |
And to him, the interesting thing was asking the questions.
link |
Yeah, try to replace me, GPT3, I dare you.
link |
Although some people think I'm a robot.
link |
You have this cool plot that shows,
link |
I just remember where economists land,
link |
where I think the X axis is the income.
link |
And then the Y axis is, I guess,
link |
aggregating the information of how replaceable the job is.
link |
Or I think there's an index.
link |
There's a suitability for machine learning index.
link |
So we have all 970 occupations on that chart.
link |
And there's scatters in all four corners
link |
have some occupations.
link |
But there is a definite pattern,
link |
which is the lower wage occupations tend to have more tasks
link |
that are suitable for machine learning, like cashiers.
link |
I mean, anyone who's gone to a supermarket or CVS
link |
knows that they not only read barcodes,
link |
but they can recognize an apple and an orange
link |
and a lot of things cashiers, humans used to be needed for.
link |
At the other end of the spectrum,
link |
there are some jobs like airline pilot
link |
that are among the highest paid in our economy,
link |
but also a lot of them are suitable for machine learning.
link |
A lot of those tasks are.
link |
And then, yeah, you mentioned economists.
link |
I couldn't help peeking at those
link |
and they're paid a fair amount,
link |
maybe not as much as some of us think they should be.
link |
But they have some tasks that are suitable
link |
for machine learning, but for now at least,
link |
most of the tasks of economists
link |
didn't end up being in that category.
link |
And I should say, I didn't like create that data.
link |
We just took the analysis and that's what came out of it.
link |
And over time, that scatter plot will be updated
link |
as the technology improves.
link |
But it was just interesting to see the pattern there.
link |
And it is a little troubling in so far
link |
as if you just take the technology as it is today,
link |
it's likely to worsen income inequality
link |
on a lot of dimensions.
link |
So on this topic of the effect of AI
link |
on our landscape of work,
link |
one of the people that have been speaking about it
link |
in the public domain, public discourse
link |
is the presidential candidate, Andrew Yang.
link |
What are your thoughts about Andrew?
link |
What are your thoughts about UBI,
link |
that universal basic income
link |
that he made one of the core ideas,
link |
by the way, he has like hundreds of ideas
link |
about like everything, it's kind of interesting.
link |
But what are your thoughts about him
link |
and what are your thoughts about UBI?
link |
Let me answer the question about his broader approach first.
link |
I mean, I just love that.
link |
He's really thoughtful, analytical.
link |
I agree with his values.
link |
So that's awesome.
link |
And he read my book and mentions it sometimes,
link |
so it makes me even more excited.
link |
And the thing that he really made the centerpiece
link |
of his campaign was UBI.
link |
And I was originally kind of a fan of it.
link |
And then as I studied it more, I became less of a fan,
link |
although I'm beginning to come back a little bit.
link |
So let me tell you a little bit of my evolution.
link |
As an economist, we have, by looking at the problem
link |
of people not having enough income and the simplest thing
link |
is, well, why don't we write them a check?
link |
But then I talked to my sociologist friends
link |
and they really convinced me that just writing a check
link |
doesn't really get at the core values.
link |
Voltaire once said that work solves three great ills,
link |
boredom, vice, and need.
link |
And you can deal with the need thing by writing a check,
link |
but people need a sense of meaning,
link |
they need something to do.
link |
And when, say, steel workers or coal miners lost their jobs
link |
and were just given checks, alcoholism, depression, divorce,
link |
all those social indicators, drug use, all went way up.
link |
People just weren't happy
link |
just sitting around collecting a check.
link |
Maybe it's part of the way they were raised.
link |
Maybe it's something innate in people
link |
that they need to feel wanted and needed.
link |
So it's not as simple as just writing people a check.
link |
You need to also give them a way to have a sense of purpose.
link |
And that was important to me.
link |
And the second thing is that, as I mentioned earlier,
link |
we are far from the end of work.
link |
I don't buy the idea that there's just like
link |
not enough work to be done.
link |
I see like our cities need to be cleaned up.
link |
And robots can't do most of that.
link |
We need to have better childcare.
link |
We need better healthcare.
link |
We need to take care of people who are mentally ill or older.
link |
We need to repair our roads.
link |
There's so much work that require at least partly,
link |
maybe entirely a human component.
link |
So rather than like write all these people off,
link |
let's find a way to repurpose them and keep them engaged.
link |
Now that said, I would like to see more buying power
link |
from people who are sort of at the bottom end
link |
The economy has been designed and evolved in a way
link |
that's I think very unfair to a lot of hardworking people.
link |
I see super hardworking people who aren't really seeing
link |
their wages grow over the past 20, 30 years,
link |
while some other people who have been super smart
link |
and or super lucky have made billions
link |
or hundreds of billions.
link |
And I don't think they need those hundreds of billions
link |
to have the right incentives to invent things.
link |
I think if you talk to almost any of them as I have,
link |
they don't think that they need an extra $10 billion
link |
to do what they're doing.
link |
Most of them probably would love to do it for only a billion
link |
or maybe for nothing.
link |
For nothing, many of them, yeah.
link |
I mean, an interesting point to make is,
link |
do we think that Bill Gates would have founded Microsoft
link |
if tax rates were 70%?
link |
Well, we know he would have because they were tax rates
link |
of 70% when he founded it.
link |
So I don't think that's as big a deterrent
link |
and we could provide more buying power to people.
link |
My own favorite tool is the Earned Income Tax Credit,
link |
which is basically a way of supplementing income
link |
of people who have jobs and giving employers
link |
an incentive to hire even more people.
link |
The minimum wage can discourage employment,
link |
but the Earned Income Tax Credit encourages employment
link |
by supplementing people's wages.
link |
If the employer can only afford to pay them $10 for a task,
link |
the rest of us kick in another five or $10
link |
and bring their wages up to 15 or 20 total.
link |
And then they have more buying power.
link |
Then entrepreneurs are thinking, how can we cater to them?
link |
How can we make products for them?
link |
And it becomes a self reinforcing system
link |
where people are better off.
link |
Ian Drang and I had a good discussion
link |
where he suggested instead of a universal basic income,
link |
he suggested, or instead of an unconditional basic income,
link |
how about a conditional basic income
link |
where the condition is you learn some new skills,
link |
we need to reskill our workforce.
link |
So let's make it easier for people to find ways
link |
to get those skills and get rewarded for doing them.
link |
And that's kind of a neat idea as well.
link |
That's really interesting.
link |
So, I mean, one of the questions,
link |
one of the dreams of UBI is that you provide
link |
some little safety net while you retrain,
link |
while you learn a new skill.
link |
But like, I think, I guess you're speaking
link |
to the intuition that that doesn't always,
link |
like there needs to be some incentive to reskill,
link |
to train, to learn a new thing.
link |
I mean, there are lots of self motivated people,
link |
but there are also people that maybe need a little guidance
link |
or help and I think it's a really hard question
link |
for someone who is losing a job in one area to know
link |
what is the new area I should be learning skills in.
link |
And we could provide a much better set of tools
link |
and platforms that maps it.
link |
Okay, here's a set of skills you already have.
link |
Here's something that's in demand.
link |
Let's create a path for you to go from where you are
link |
to where you need to be.
link |
So I'm a total, how do I put it nicely about myself?
link |
I'm totally clueless about the economy.
link |
It's not totally true, but pretty good approximation.
link |
If you were to try to fix our tax system
link |
and, or maybe from another side,
link |
if there's fundamental problems in taxation
link |
or some fundamental problems about our economy,
link |
what would you try to fix?
link |
What would you try to speak to?
link |
You know, I definitely think our whole tax system,
link |
our political and economic system has gotten more
link |
and more screwed up over the past 20, 30 years.
link |
I don't think it's that hard to make headway
link |
I don't think we need to totally reinvent stuff.
link |
A lot of it is what I've been elsewhere with Andy
link |
and others called economics 101.
link |
You know, there's just some basic principles
link |
that have worked really well in the 20th century
link |
that we sort of forgot, you know,
link |
in terms of investing in education,
link |
investing in infrastructure, welcoming immigrants,
link |
having a tax system that was more progressive and fair.
link |
At one point, tax rates were on top incomes
link |
were significantly higher.
link |
And they've come down a lot to the point where
link |
in many cases they're lower now
link |
than they are for poorer people.
link |
So, and we could do things like earned income tax credit
link |
to get a little more wonky.
link |
I'd like to see more Pigouvian taxes.
link |
What that means is you tax things that are bad
link |
instead of things that are good.
link |
So right now we tax labor, we tax capital
link |
and which is unfortunate
link |
because one of the basic principles of economics
link |
if you tax something, you tend to get less of it.
link |
So, you know, right now there's still work to be done
link |
and still capital to be invested in.
link |
But instead we should be taxing things like pollution
link |
And if we did that, we would have less pollution.
link |
So a carbon tax is, you know,
link |
almost every economist would say it's a no brainer
link |
whether they're Republican or Democrat,
link |
Greg Mankiw who is head of George Bush's
link |
Council of Economic Advisers or Dick Schmollensie
link |
who is another Republican economist agree.
link |
And of course a lot of Democratic economists agree as well.
link |
If we taxed carbon,
link |
we could raise hundreds of billions of dollars.
link |
We could take that money and redistribute it
link |
through an earned income tax credit or other things
link |
so that overall our tax system would become more progressive.
link |
We could tax congestion.
link |
One of the things that kills me as an economist
link |
is every time I sit in a traffic jam,
link |
I know that it's completely unnecessary.
link |
This is complete wasted time.
link |
You just visualize the cost and productivity.
link |
Exactly, because they are taking costs for me
link |
and all the people around me.
link |
And if they charged a congestion tax,
link |
they would take that same amount of money
link |
and people would, it would streamline the roads.
link |
Like when you're in Singapore, the traffic just flows
link |
because they have a congestion tax.
link |
They listened to economists.
link |
They invited me and others to go talk to them.
link |
And then I'd still be paying,
link |
I'd be paying a congestion tax instead of paying in my time,
link |
but that money would now be available for healthcare,
link |
be available for infrastructure,
link |
or be available just to give to people
link |
so they could buy food or whatever.
link |
So it's just, it saddens me when you sit,
link |
when you're sitting in a traffic jam,
link |
it's like taxing me and then taking that money
link |
and dumping it in the ocean, just like destroying it.
link |
So there are a lot of things like that
link |
that economists, and I'm not,
link |
I'm not like doing anything radical here.
link |
Most, you know, good economists would,
link |
I probably agree with me point by point on these things.
link |
And we could do those things
link |
and our whole economy would become much more efficient.
link |
It'd become fairer, invest in R&D and research,
link |
which is close to a free lunch is what we have.
link |
My erstwhile MIT colleague, Bob Solla,
link |
got the Nobel Prize, not yesterday, but 30 years ago,
link |
for describing that most improvements
link |
in living standards come from tech progress.
link |
And Paul Romer later got a Nobel Prize
link |
for noting that investments in R&D and human capital
link |
can speed the rate of tech progress.
link |
So if we do that, then we'll be healthier and wealthier.
link |
Yeah, from an economics perspective,
link |
I remember taking an undergrad econ,
link |
you mentioned econ 101.
link |
It seemed from all the plots I saw
link |
that R&D is an obvious, as close to free lunch as we have,
link |
it seemed like obvious that we should do more research.
link |
Like what, what, like, there's no.
link |
Well, we should do basic research.
link |
I mean, so let me just be clear.
link |
It'd be great if everybody did more research
link |
and I would make this issue
link |
between applied development versus basic research.
link |
So applied development, like, you know,
link |
how do we get this self driving car, you know,
link |
feature to work better in the Tesla?
link |
That's great for private companies
link |
because they can capture the value from that.
link |
If they make a better self driving car system,
link |
they can sell cars that are more valuable
link |
and then make money.
link |
So there's an incentive that there's not a big problem there
link |
and smart companies, Amazon, Tesla,
link |
and others are investing in it.
link |
The problem is with basic research,
link |
like coming up with core basic ideas,
link |
whether it's in nuclear fusion
link |
or artificial intelligence or biotech.
link |
There, if someone invents something,
link |
it's very hard for them to capture the benefits from it.
link |
It's shared by everybody, which is great in a way,
link |
but it means that they're not gonna have the incentives
link |
to put as much effort into it.
link |
There you need, it's a classic public good.
link |
There you need the government to be involved in it.
link |
And the US government used to be investing much more in R&D,
link |
but we have slashed that part of the government
link |
really foolishly and we're all poorer,
link |
significantly poorer as a result.
link |
Growth rates are down.
link |
We're not having the kind of scientific progress
link |
It's been sort of a short term eating the seed corn,
link |
whatever metaphor you wanna use
link |
where people grab some money, put it in their pockets today,
link |
but five, 10, 20 years later,
link |
they're a lot poorer than they otherwise would have been.
link |
So we're living through a pandemic right now,
link |
globally in the United States.
link |
From an economics perspective,
link |
how do you think this pandemic will change the world?
link |
It's been remarkable.
link |
And it's horrible how many people have suffered,
link |
the amount of death, the economic destruction.
link |
It's also striking just the amount of change in work
link |
In the last 20 weeks, I've seen more change
link |
than there were in the previous 20 years.
link |
There's been nothing like it
link |
since probably the World War II mobilization
link |
in terms of reorganizing our economy.
link |
The most obvious one is the shift to remote work.
link |
And I and many other people stopped going into the office
link |
and teaching my students in person.
link |
I did a study on this with a bunch of colleagues
link |
at MIT and elsewhere.
link |
And what we found was that before the pandemic,
link |
in the beginning of 2020, about one in six,
link |
a little over 15% of Americans were working remotely.
link |
When the pandemic hit, that grew steadily and hit 50%,
link |
roughly half of Americans working at home.
link |
So a complete transformation.
link |
And of course, it wasn't even,
link |
it wasn't like everybody did it.
link |
If you're an information worker, professional,
link |
if you work mainly with data,
link |
then you're much more likely to work at home.
link |
If you're a manufacturing worker,
link |
working with other people or physical things,
link |
then it wasn't so easy to work at home.
link |
And instead, those people were much more likely
link |
to become laid off or unemployed.
link |
So it's been something that's had very disparate effects
link |
on different parts of the workforce.
link |
Do you think it's gonna be sticky in a sense
link |
that after vaccine comes out and the economy reopens,
link |
do you think remote work will continue?
link |
That's a great question.
link |
My hypothesis is yes, a lot of it will.
link |
Of course, some of it will go back,
link |
but a surprising amount of it will stay.
link |
I personally, for instance, I moved my seminars,
link |
my academic seminars to Zoom,
link |
and I was surprised how well it worked.
link |
Yeah, I mean, obviously we were able to reach
link |
a much broader audience.
link |
So we have people tuning in from Europe
link |
and other countries,
link |
just all over the United States for that matter.
link |
I also actually found that it would,
link |
in many ways, is more egalitarian.
link |
We use the chat feature and other tools,
link |
and grad students and others who might've been
link |
a little shy about speaking up,
link |
we now kind of have more of ability for lots of voices.
link |
And they're answering each other's questions,
link |
so you kind of get parallel.
link |
Like if someone had some question about some of the data
link |
or a reference or whatever,
link |
then someone else in the chat would answer it.
link |
And the whole thing just became like a higher bandwidth,
link |
higher quality thing.
link |
So I thought that was kind of interesting.
link |
I think a lot of people are discovering that these tools
link |
that thanks to technologists have been developed
link |
over the past decade,
link |
they're a lot more powerful than we thought.
link |
I mean, all the terrible things we've seen with COVID
link |
and the real failure of many of our institutions
link |
that I thought would work better.
link |
One area that's been a bright spot is our technologies.
link |
Bandwidth has held up pretty well,
link |
and all of our email and other tools
link |
have just scaled up kind of gracefully.
link |
So that's been a plus.
link |
Economists call this question
link |
of whether it'll go back a hysteresis.
link |
The question is like when you boil an egg
link |
after it gets cold again, it stays hard.
link |
And I think that we're gonna have a fair amount
link |
of hysteresis in the economy.
link |
We're gonna move to this new,
link |
we have moved to a new remote work system,
link |
and it's not gonna snap all the way back
link |
to where it was before.
link |
One of the things that worries me is that the people
link |
with lots of followers on Twitter and people with voices,
link |
people that can, voices that can be magnified by reporters
link |
and all that kind of stuff are the people
link |
that fall into this category
link |
that we were referring to just now
link |
where they can still function
link |
and be successful with remote work.
link |
And then there is a kind of quiet suffering
link |
of what feels like millions of people
link |
whose jobs are disturbed profoundly by this pandemic,
link |
but they don't have many followers on Twitter.
link |
What do we, and again, I apologize,
link |
but I've been reading the rise and fall of the Third Reich
link |
and there's a connection to the depression
link |
on the American side.
link |
There's a deep, complicated connection
link |
to how suffering can turn into forces
link |
that potentially change the world in destructive ways.
link |
So like it's something I worry about is like,
link |
what is this suffering going to materialize itself
link |
in five, 10 years?
link |
Is that something you worry about, think about?
link |
It's like the center of what I worry about.
link |
And let me break it down to two parts.
link |
There's a moral and ethical aspect to it.
link |
We need to relieve this suffering.
link |
I mean, I'm sure the values of, I think most Americans,
link |
we like to see shared prosperity
link |
or most people on the planet.
link |
And we would like to see people not falling behind
link |
and they have fallen behind, not just due to COVID,
link |
but in the previous couple of decades,
link |
median income has barely moved,
link |
depending on how you measure it.
link |
And the incomes of the top 1% have skyrocketed.
link |
And part of that is due to the ways technology has been used.
link |
Part of this been due to, frankly, our political system
link |
has continually shifted more wealth into those people
link |
who have the powerful interest.
link |
So there's just, I think, a moral imperative
link |
to do a better job.
link |
And ultimately, we're all gonna be wealthier
link |
if more people can contribute,
link |
more people have the wherewithal.
link |
But the second thing is that there's a real political risk.
link |
I'm not a political scientist,
link |
but you don't have to be one, I think,
link |
to see how a lot of people are really upset
link |
with they're getting a raw deal
link |
and they want to smash the system in different ways,
link |
And now I think there are a lot of people
link |
who are looking at the political system
link |
and they feel like it's not working for them
link |
and they just wanna do something radical.
link |
Unfortunately, demagogues have harnessed that
link |
in a way that is pretty destructive to the country.
link |
And an analogy I see is what happened with trade.
link |
Almost every economist thinks that free trade
link |
is a good thing, that when two people voluntarily exchange
link |
almost by definition, they're both better off
link |
if it's voluntary.
link |
And so generally, trade is a good thing.
link |
But they also recognize that trade can lead
link |
to uneven effects, that there can be winners and losers
link |
in some of the people who didn't have the skills
link |
to compete with somebody else or didn't have other assets.
link |
And so trade can shift prices
link |
in ways that are averse to some people.
link |
So there's a formula that economists have,
link |
which is that you have free trade,
link |
but then you compensate the people who are hurt
link |
and free trade makes the pie bigger.
link |
And since the pie is bigger,
link |
it's possible for everyone to be better off.
link |
You can make the winners better off,
link |
but you can also compensate those who don't win.
link |
And so they end up being better off as well.
link |
What happened was that we didn't fulfill that promise.
link |
We did have some more increased free trade
link |
in the 80s and 90s, but we didn't compensate the people
link |
And so they felt like the people in power
link |
reneged on the bargain, and I think they did.
link |
And so then there's a backlash against trade.
link |
And now both political parties,
link |
but especially Trump and company,
link |
have really pushed back against free trade.
link |
Ultimately, that's bad for the country.
link |
Ultimately, that's bad for living standards.
link |
But in a way I can understand
link |
that people felt they were betrayed.
link |
Technology has a lot of similar characteristics.
link |
Technology can make us all better off.
link |
It makes the pie bigger.
link |
It creates wealth and health, but it can also be uneven.
link |
Not everyone automatically benefits.
link |
It's possible for some people,
link |
even a majority of people to get left behind
link |
while a small group benefits.
link |
What most economists would say,
link |
well, let's make the pie bigger,
link |
but let's make sure we adjust the system
link |
so we compensate the people who are hurt.
link |
And since the pie is bigger,
link |
we can make the rich richer,
link |
we can make the middle class richer,
link |
we can make the poor richer.
link |
Mathematically, everyone could be better off.
link |
But again, we're not doing that.
link |
And again, people are saying this isn't working for us.
link |
And again, instead of fixing the distribution,
link |
a lot of people are beginning to say,
link |
hey, technology sucks, we've got to stop it.
link |
Let's throw rocks at the Google bus.
link |
And there were the Luddites almost exactly 200 years ago
link |
who smashed the looms and the spinning machines
link |
because they felt like those machines weren't helping them.
link |
We have a real imperative,
link |
not just to do the morally right thing,
link |
but to do the thing that is gonna save the country,
link |
which is make sure that we create
link |
not just prosperity, but shared prosperity.
link |
So you've been at MIT for over 30 years, I think.
link |
Don't tell anyone how old I am.
link |
Yeah, no, that's true, that's true.
link |
And you're now moved to Stanford.
link |
I'm gonna try not to say anything
link |
about how great MIT is.
link |
What's that move been like?
link |
What, it's East Coast to West Coast?
link |
Well, MIT is great.
link |
MIT has been very good to me.
link |
It continues to be very good to me.
link |
It's an amazing place.
link |
I continue to have so many amazing friends
link |
and colleagues there.
link |
I'm very fortunate to have been able
link |
to spend a lot of time at MIT.
link |
Stanford's also amazing.
link |
And part of what attracted me out here
link |
was not just the weather, but also Silicon Valley,
link |
let's face it, is really more of the epicenter
link |
of the technological revolution.
link |
And I wanna be close to the people
link |
who are inventing AI and elsewhere.
link |
A lot of it is being invested at MIT for that matter
link |
in Europe and China and elsewhere, in Nia.
link |
But being a little closer to some of the key technologists
link |
was something that was important to me.
link |
And it may be shallow,
link |
but I also do enjoy the good weather.
link |
And I felt a little ripped off
link |
when I came here a couple of months ago.
link |
And immediately there are the fires
link |
and my eyes were burning, the sky was orange
link |
and there's the heat waves.
link |
And so it wasn't exactly what I've been promised,
link |
but fingers crossed it'll get back to better.
link |
But maybe on a brief aside,
link |
there's been some criticism of academia
link |
and universities and different avenues.
link |
And I, as a person who's gotten to enjoy universities
link |
from the pure playground of ideas that it can be,
link |
always kind of try to find the words
link |
to tell people that these are magical places.
link |
Is there something that you can speak to
link |
that is beautiful or powerful about universities?
link |
I mean, first off, I mean,
link |
economists have this concept called revealed preference.
link |
You can ask people what they say
link |
or you can watch what they do.
link |
And so obviously by reveal preferences, I love academia.
link |
I could be doing lots of other things,
link |
but it's something I enjoy a lot.
link |
And I think the word magical is exactly right.
link |
At least it is for me.
link |
I do what I love, you know,
link |
hopefully my Dean won't be listening,
link |
but I would do this for free.
link |
You know, it's just what I like to do.
link |
I like to do research.
link |
I love to have conversations like this with you
link |
and with my students, with my fellow colleagues.
link |
I love being around the smartest people I can find
link |
and learning something from them
link |
and having them challenge me.
link |
And that just gives me joy.
link |
And every day I find something new and exciting to work on.
link |
And a university environment is really filled
link |
with other people who feel that way.
link |
And so I feel very fortunate to be part of it.
link |
And I'm lucky that I'm in a society
link |
where I can actually get paid for it
link |
and put food on the table
link |
while doing the stuff that I really love.
link |
And I hope someday everybody can have jobs
link |
that are like that.
link |
And I appreciate that it's not necessarily easy
link |
for everybody to have a job that they both love
link |
and also they get paid for.
link |
So there are things that don't go well in academia,
link |
but by and large, I think it's a kind of, you know,
link |
kinder, gentler version of a lot of the world.
link |
You know, we sort of cut each other a little slack
link |
on things like, you know, on just a lot of things.
link |
You know, of course there's harsh debates
link |
and discussions about things
link |
and some petty politics here and there.
link |
I personally, I try to stay away
link |
from most of that sort of politics.
link |
It's not my thing.
link |
And so it doesn't affect me most of the time,
link |
sometimes a little bit, maybe.
link |
But, you know, being able to pull together something,
link |
we have the digital economy lab.
link |
We've got all these brilliant grad students
link |
and undergraduates and postdocs
link |
that are just doing stuff that I learned from.
link |
And every one of them has some aspect
link |
of what they're doing that's just,
link |
I couldn't even understand.
link |
It's like way, way more brilliant.
link |
And that's really, to me, actually I really enjoy that,
link |
being in a room with lots of other smart people.
link |
And Stanford has made it very easy to attract,
link |
you know, those people.
link |
I just, you know, say I'm gonna do a seminar, whatever,
link |
and the people come, they come and wanna work with me.
link |
We get funding, we get data sets,
link |
and it's come together real nicely.
link |
And the rest is just fun.
link |
And we feel like we're working on important problems,
link |
you know, and we're doing things that, you know,
link |
I think are first order in terms of what's important
link |
in the world, and that's very satisfying to me.
link |
Maybe a bit of a fun question.
link |
What three books, technical, fiction, philosophical,
link |
you've enjoyed, had a big, big impact in your life?
link |
Well, I guess I go back to like my teen years,
link |
and, you know, I read Sid Arthur,
link |
which is a philosophical book,
link |
and kind of helps keep me centered.
link |
Yeah, by Herman Hesse, exactly.
link |
Don't get too wrapped up in material things
link |
or other things, and just sort of, you know,
link |
try to find peace on things.
link |
A book that actually influenced me a lot
link |
in terms of my career was called
link |
The Worldly Philosophers by Robert Halbrenner.
link |
It's actually about economists.
link |
It goes through a series of different,
link |
it's written in a very lively form,
link |
and it probably sounds boring,
link |
but it did describe whether it's Adam Smith
link |
or Karl Marx or John Maynard Keynes,
link |
and each of them sort of what their key insights were,
link |
but also kind of their personalities,
link |
and I think that's one of the reasons
link |
I became an economist was just understanding
link |
how they grapple with the big questions of the world.
link |
So would you recommend it as a good whirlwind overview
link |
of the history of economics?
link |
Yeah, yeah, I think that's exactly right.
link |
It kind of takes you through the different things,
link |
and so you can understand how they reach,
link |
thinking some of the strengths and weaknesses.
link |
I mean, it probably is a little out of date now.
link |
It needs to be updated a bit,
link |
but you could at least look through
link |
the first couple hundred years of economics,
link |
which is not a bad place to start.
link |
More recently, I mean, a book I really enjoyed
link |
is by my friend and colleague, Max Tegmark,
link |
You should have him on your podcast if you haven't already.
link |
He was episode number one.
link |
And he's back, he'll be back, he'll be back soon.
link |
Yeah, no, he's terrific.
link |
I love the way his brain works,
link |
and he makes you think about profound things.
link |
He's got such a joyful approach to life,
link |
and so that's been a great book,
link |
and I learn a lot from it, I think everybody,
link |
but he explains it in a way, even though he's so brilliant,
link |
that everyone can understand, that I can understand.
link |
That's three, but let me mention maybe one or two others.
link |
I mean, I recently read More From Less
link |
by my sometimes coauthor, Andrew McAfee.
link |
It made me optimistic about how we can continue
link |
to have rising living standards
link |
while living more lightly on the planet.
link |
In fact, because of higher living standards,
link |
because of technology,
link |
because of digitization that I mentioned,
link |
we don't have to have as big an impact on the planet,
link |
and that's a great story to tell,
link |
and he documents it very carefully.
link |
You know, a personal kind of self help book
link |
that I found kind of useful, People, is Atomic Habits.
link |
I think it's, what's his name, James Clear.
link |
Yeah, James Clear.
link |
He's just, yeah, it's a good name,
link |
because he writes very clearly,
link |
and you know, most of the sentences I read in that book,
link |
I was like, yeah, I know that,
link |
but it just really helps to have somebody like remind you
link |
and tell you and kind of just reinforce it, and it's helpful.
link |
So build habits in your life that you hope to have,
link |
that have a positive impact,
link |
and don't have to make it big things.
link |
It could be just tiny little.
link |
Exactly, I mean, the word atomic,
link |
it's a little bit of a pun, I think he says.
link |
You know, one, atomic means they're really small.
link |
You take these little things, but also like atomic power,
link |
can have like, you know, big impact.
link |
That's funny, yeah.
link |
The biggest ridiculous question,
link |
especially to ask an economist, but also a human being,
link |
what's the meaning of life?
link |
I hope you've gotten the answer to that from somebody else.
link |
I think we're all still working on that one, but what is it?
link |
You know, I actually learned a lot from my son, Luke,
link |
and he's 19 now, but he's always loved philosophy,
link |
and he reads way more sophisticated philosophy than I do.
link |
I went and took him to Oxford,
link |
and he spent the whole time like pulling
link |
all these obscure books down and reading them.
link |
And a couple of years ago, we had this argument,
link |
and he was trying to convince me that hedonism
link |
was the ultimate, you know, meaning of life,
link |
just pleasure seeking, and...
link |
Well, how old was he at the time?
link |
But he made a really good like intellectual argument
link |
for it too, and you know,
link |
but you know, it just didn't strike me as right.
link |
And I think that, you know, while I am kind of a utilitarian,
link |
like, you know, I do think we should do the grace,
link |
good for the grace number, that's just too shallow.
link |
And I think I've convinced myself that real happiness
link |
doesn't come from seeking pleasure.
link |
It's kind of a little, it's ironic.
link |
Like if you really focus on being happy,
link |
I think it doesn't work.
link |
You gotta like be doing something bigger.
link |
I think the analogy I sometimes use is, you know,
link |
when you look at a dim star in the sky,
link |
if you look right at it, it kind of disappears,
link |
but you have to look a little to the side,
link |
and then the parts of your retina
link |
that are better at absorbing light,
link |
you know, can pick it up better.
link |
It's the same thing with happiness.
link |
I think you need to sort of find something, other goal,
link |
something, some meaning in life,
link |
and that ultimately makes you happier
link |
than if you go squarely at just pleasure.
link |
And so for me, you know, the kind of research I do
link |
that I think is trying to change the world,
link |
make the world a better place,
link |
and I'm not like an evolutionary psychologist,
link |
but my guess is that our brains are wired,
link |
not just for pleasure, but we're social animals,
link |
and we're wired to like help others.
link |
And ultimately, you know,
link |
that's something that's really deeply rooted in our psyche.
link |
And if we do help others, if we do,
link |
or at least feel like we're helping others,
link |
you know, our reward systems kick in,
link |
and we end up being more deeply satisfied
link |
than if we just do something selfish and shallow.
link |
I don't think there's a better way to end it, Eric.
link |
You were one of the people when I first showed up at MIT,
link |
that made me proud to be at MIT.
link |
So it's so sad that you're now at Stanford,
link |
but I'm sure you'll do wonderful things at Stanford as well.
link |
I can't wait till future books,
link |
and people should definitely read your other books.
link |
Well, thank you so much.
link |
And I think we're all part of the invisible college,
link |
You know, we're all part of this intellectual
link |
and human community where we all can learn from each other.
link |
It doesn't really matter physically
link |
where we are so much anymore.
link |
Thanks for talking today.
link |
Thanks for listening to this conversation
link |
with Eric Brynjolfsson.
link |
And thank you to our sponsors.
link |
Vincero Watches, the maker of classy,
link |
well performing watches.
link |
Fort Sigmatic, the maker of delicious mushroom coffee.
link |
ExpressVPN, the VPN I've used for many years
link |
to protect my privacy on the internet.
link |
And CashApp, the app I use to send money to friends.
link |
Please check out these sponsors in the description
link |
to get a discount and to support this podcast.
link |
If you enjoy this thing, subscribe on YouTube.
link |
Review it with five stars on Apple Podcast,
link |
follow on Spotify, support on Patreon,
link |
or connect with me on Twitter at Lex Friedman.
link |
And now, let me leave you with some words
link |
from Albert Einstein.
link |
It has become appallingly obvious
link |
that our technology has exceeded our humanity.
link |
Thank you for listening and hope to see you next time.