3 Takeaways - What Does the Work of The Future Look Like? With MIT Professor David Autor (#116)
Episode Date: October 25, 2022Why will tech and automation never lead to the demise of human work? What qualifies as “good” work? What role will robots and AI play in the fast-approaching future? David Autor, MIT professor and... co-chair of the MIT Task Force on The Work of The Future, provides answers in this riveting and enlightening conversation.
Transcript
Discussion (0)
Welcome to the Three Takeaways podcast, which features short, memorable conversations with the world's best thinkers, business leaders, writers, politicians, scientists, and other newsmakers.
Each episode ends with the three key takeaways that person has learned over their lives and their careers.
And now your host and board member of schools at Harvard, Princeton, and Columbia, Lynn Thoman.
Hi, everyone. It's Lynn Thoman. Hi, everyone.
It's Lynn Thoman.
Welcome to another episode.
Today, I'm excited to be with MIT professor David Otter to find out about the future of work.
It's such an important question.
Will machines increasingly do our work for us,
and will they crush employment and jobs?
And what does the work of the future actually look like?
David Otter is just the person to ask.
He's an award-winning MIT professor and co-chair of the MIT Task Force on the Work of the Future.
Welcome, David, and thanks so much for our conversation today.
Thank you for inviting me.
My pleasure.
Many people thought computers would replace workers doing repetitive mundane tasks everywhere, including in offices, factory floors and outdoors.
But if that were the case, why are there so many jobs now?
Computers have done a lot of those replacing things. They certainly have replaced people in offices.
They certainly have replaced people on factory floors.
But there are two, actually three major reasons why that
hasn't had the net effect of eliminated employment. One is that actually those things have the free
time and they create income, those same process of replacement, and that creates wealth and that
causes people to consume. And because people are insatiable, that creates more employment onto
itself. We constantly create work for each other just by our consumption.
A second, probably at least equally important reason, maybe even more important is that many of those technologies are not simply replacing technologies. They're tools that augment us.
If you said to a roofer, I'm going to take away your pneumatic nail gun because it's just replacing
your work. They'd say, you're crazy. I can't get anything done if I did that. I couldn't do these
jobs, right? Many tools compliment us, right? So imagine if we took away your computer, would you
say, oh, now my job is more valuable? Or would you say, wow, now I'm not really able to actually
accomplish any of the things I'm setting out to do. Tools often compliment us. And even as they
take away the mundane part of our tasks, they actually allow us to focus our expertise,
our judgment, our creativity on the
thing where we really have comparative advantage. A third reason is actually we're constantly
creating new activities. It's not just that we do the same things more efficiently with technology.
We create new types of work. Think of the medical specialties, think of the technology skills,
but even all kinds of new services that didn't exist, whether they're in entertainment, the recreation, or care for young people, or care for seniors, therapists, and
coaches. And those things reflect rising incomes, they reflect changing tastes, they reflect
changing demographics, and they reflect overall creativity. So it's not simply that we do more of
the same more efficiently with machines and machines do some of them. We are creating new
tasks. In fact, my recent work with co-authors Anna Solomons and Carolyn Chin and Brian Segmiller,
we estimate that about 61% of the jobs that people do in 2018 essentially did not exist in 1940.
A lot of the work that we do is new in the last century. And speaking of the last century, 100 years ago, 40% of Americans
worked on farms, and today only about 2% do. Of course, it's phenomenal that 2% of the labor
force can produce enough food for a country of over 300 million people. But what happened to
all those farmers whose jobs essentially disappeared? Yeah, absolutely. It's an incredible
technological triumph. And this reflects improvements in irrigation, in genetics,
in fertilization, in mechanization. So yeah, a couple million people can feed a country of 300
million and we're a huge food exporter. Where did all those people go? Well, it was a generational
change and it did happen over the course of decades that we had this major decline.
But a couple of things were happening simultaneously that were very beneficial.
One was the rise of industry, this incredible growth of factory work.
It took a lot of people without high levels of formal education and made them highly productive in building stuff.
So, right, the rise in mass production accompanied the decline of agriculture.
And in some sense, that was partly coincidence.
It was not that we created one to deal with the problem of the other,
but we were very fortunate that those things happened simultaneously.
A second thing is that the citizens of the US farm states made a very forward-looking decision around the late 19th and early 20th century, which was to mandate that all their kids go to high
school and stay in school until the age of 18. And that was a crazy decision at the time. It seemed crazy because
it was expensive. Not only did you have to hire teachers and build buildings and buy books,
but those kids couldn't work on the farm during those years. So that was the biggest cost.
But also it seemed like overkill. Why do all these people need to have this elite skill? Why do they
need to be both literate and numerate? Isn't that too much? And the rest in Europe thought we were kind of
nuts to do this. It turned out that it gave the US the most flexible, the most productive,
the most skilled workforce in the world at that time. It allowed us to successfully ramp up
production during the Second World War and make these transitions. So education was an incredibly
important part of this. And if we hadn't educated ourselves and yet somehow hypothetically have the same technologies
we have now, we wouldn't be good at them because almost all jobs at this point require literacy
and numeracy. And without those things, people would have sound characters and strong backs,
but they wouldn't actually be productive in a high-tech economy such as we have.
So that was a successful transition.
Let's talk about how jobs have changed.
Can you take a couple of different jobs and explain how each job has changed from about
20 years ago to today?
The most, I think, numerically important one is office clerical work.
There used to be a lot more office clerical
workers who did phone answering, filing, typing, duplicating, a lot of what you might think of as
routine information processing. I'm using routine in a very specific sense. I don't mean routine in
the sense of mundane. I mean routine in the sense of follows a well-understood set of rules and
procedures, a routine. And those tasks have proved highly amenable to computerization.
Well before artificial intelligence, right? We were creating word processing. We were creating
databases and spreadsheets and phone trees. And so a lot of the paperwork of office work
has been reduced. A lot of people now type their own documents and process them. They don't need a third party. And this has dramatically decreased employment in office
administrative work. Now, the jobs that remain are more skilled and more demanding. What does
an office administrator do now? Well, they coordinate events. They manage travel. They
deal with those dreaded receipts. They help proof papers.
And so they are basically often orchestrators of more complex activities.
So in general, they're higher educated, they're higher paid, but there are many fewer of them.
So it's a more skilled activity. And this is what we see in a lot of white collar work that is being subject to automation.
Artificial intelligence is accelerating this process.
Let's take another to counterbalance that one.
Think of the case of London cabbies.
So London taxi drivers,
I know we're not in the UK at the moment,
but to become a London taxi driver,
you had to memorize all of the streets in London.
It was an incredible feat of memory
to be certified to be a London cab driver.
They wouldn't have to consult a map.
They would get you there.
And it had a lot of prestige for that reason as well. And of course, they have been substantially
supplanted by Uber and Lyft and all the ride hailing apps. So what have those done? Well,
they've had real pluses and minuses. They've taken a lot of the expertise out of the work.
There's nothing that a London cabbie could do that Waze can't do at least as well,
at least in terms of routing. And so they've opened up the occupation to many more people,
which is good, lowered the cost, which is good. They made it more convenient for customers,
which is good. They've also created a type of work that is unusual, which is ride-hailing.
It's a job you can turn on and off like water, right? You can say, I want to work these hours,
not that hours. This looks like high time, no time. So that's good. However, they've also
really commodified the work. They've taken the expertise out of it. Basically, the computer does all the hard thinking around driving and navigation, and
the person is left to deal with what's ironically a task that's very hard to computerize, but
not at all challenging for people.
We do not yet have really good self-driving technologies, but my 16-year-old kid can drive
a car just fine without a lot of thinking.
It just uses onboard equipment that's built into people.
There's a lot of things that we know how to do tacitly that it's actually hard to explain
explicitly.
They're hard to computerize.
So the philosopher Michael Polanyi once said, we know more than we can tell.
There are many things that we do that we don't know how we do them.
And that has historically been a barrier to automation.
And this is an irony of a lot of the computerized world,
that there's sort of two broad sets of activities that have not so far been amenable to automation.
One is a lot of the professional, technical, managerial, creative thinking, analytical,
even care work that we do. Go down a level and you go to the office clerical and make production
tasks. Those are middle skilled work and those have been automated because they follow these well-understood rules and procedures,
these routines. Now, say driving a car, cleaning a room, being security, cooking a meal, home health
aides, these jobs have proven very hard to automate. And you might think, oh, therefore
they would pay well, but that's not true. And the reason is because the skill sets required
are highly generic. Most people can do them, even though they're extremely hard to automate at present, they eventually will be. The number of
people who can do them is enormous. These are like life and death occupations, but they pay poorly.
And part of the reason they pay poorly is because the skills to do them are not scarce. They don't
require much what we call expertise. How do you think artificial intelligence will change work,
productivity, and also the way people lead their lives?
First of all, I should say that artificial intelligence is extremely promising and
extremely uncertain. Not only is it hard to say what it can do, it's also very hard to say what
it can't do. And so it's not easy to make confident forecasts about what AI will do. Computers
accomplishing routine tasks. This is something we've understood well for 20 years, right? We
understand procedural programming and all the steps required. And that's why we could say,
oh, this thing will be computerized. Chess games will be computerized long before hotel room
cleaning is computerized because chess is a closed-end game. We know exactly what the rules
are, literally. But AI is different from that because it's actually mysterious to the people
who create it. Even it's a mysterious how well it works and also can't explain itself to us.
So you can't ask an AI, oh, how did you figure that out? The way that information is in some
sense processed is totally foreign to the way we understand it, right? It's just a bunch
of billions of weights and connections and so on. So it's very difficult to say what AI will and
will not be able to do well. A further point to add to that, I think is really central. It's a
very broadly applicable technology, a general technology. So part of what it will do or can
do depends on what we want it to do and what things we prioritize to use it for.
And in many ways, AI research is driven by the model of let's replicate human capabilities.
That's always been the holy grail. But if you think about it, we already have human capabilities.
Like replicating them isn't so useful. It'd be better to actually do things we couldn't do with
humans with AI. But it will make many more things, quote, routine in some sense that we don't think
of as routine now.
Will AI mean that there's less of a premium on knowledge?
It will certainly mean there's less of a premium on factual knowledge.
And we already see this.
In fact, there's research showing that basically people are less good at rote memorization
and know fewer, just pure facts
than they used to. And they much more now focus in terms of their own skill development is seen
on IQ tests on basically building analytical skills, right? And that makes total sense
because there was a time when you couldn't just look on your phone and figure out who was the
president, how many ounces in a pound, how many grams in an ounce, et cetera? So it will be the case, yes, that the premium to just knowing stuff will decline, but the
premium to be able to synthesize and use information well will rise.
It's sort of like I say, well, shouldn't my kids learn math the way they used to?
I don't actually think there's a big value to learning times tables anymore.
However, there's enormous value to learning how to work with data, how to take a mass
of information and glean conclusions, analytically supported conclusions from it or work with
that.
Our world is awash in data and interpreting data is very different from just crunching
information, right?
We can make all the calculations we want, but that doesn't tell us what to think.
To think we have to have a hypothesis about, well, what might be going on here?
We have to have data that we can use to evaluate it, support it or refute it. And that is enormously important skill, and not just for a
research scientist, for anyone navigating in a world that's awash with information and misinformation.
You've talked a lot about higher order skills. What's happened to low-skill jobs,
jobs for those without college degrees? This is the big problem. It's this quality
problem, not the quantity problem. As the middle has hollowed out, as we've lost clerical
administrative support, production and operative positions, increasingly people without college
degrees or the majority of adults, a substantial majority, have been relegated downward into
occupations using generic skill sets, food service, cleaning, security.
And that's really a problem. So it's not that they can't find work, but that work doesn't pay well.
And moreover, it doesn't lead to rising productivity over the life cycle. You don't
get that much, like once you're a cashier, you might learn a lot in the first one or two weeks
or a month, but then you hit peak, you're not getting better. So you're always in competition
with the next person who walks in the door and it doesn't provide a career ladder
into a better job. So that's really the problem that we face. And that's why I think it's so
critical thing about how do we restore expert work for people who don't have huge amounts of
formal training? How can you make them more productive? It's not an easy problem, but it's not a kind of impossible
philosophical problem. So we want to be able to restore expert work for people without high
levels of education because that's where good livelihoods come from. They come from knowing
how to do something that's valuable that not everybody else can do. The job quality problem
is an enormous one. I think it's a problem for the distribution of income. It's a problem for our politics.
It's a problem for a healthy democratic society, for many, many people to feel like their horizons
have been shortened.
The type of livelihoods that they used to do are less valuable.
And the opportunities are only run through what is perceived as a lead institution, which
is higher education.
David, what is a good job or what you call a good-ish job?
Not to be crass, but the first thing that matters is compensation.
Does a job allow you to meet a decent standard of living, meaning you can take care of your
family, your kids can live in a safe neighborhood, go to a good school?
And all else equal, setting aside intrinsic
worth, setting aside good treatment, people are happier when they're paid more, especially when
we go from very low to reasonable standard of living. So all else equal, we want everyone
to have a job that allowed them to have a reasonable standard of living in a modern society.
But on top of that, what makes work good is that it uses some expertise,
some skill that you have, whether it's in care, whether it's in design, whether it's in construction,
whether it's in research, whether it's in cooking, that you have that's meaningful and that's
distinct. And that distinction matters because it means that you are not easily replaceable,
you are not generic. So you will intrinsically be treated better by your employer. If you can do something that others
can't do, you will not be immediately in competition with the next 16-year-old kid who
shows up to say, oh, I could do that. And along with that expertise will come social esteem
because you're the person who's recognized and good at this. It doesn't have to be you're the
only best one in the world. It's not that you have to be on the Jimmy Fallon show to talk about what you do. It simply means
you're good at it. People know it's valuable. That commands respect. Of course, on top of that,
we want people to be doing work that they consider to be socially valuable, or at least not socially
destructive. Although some people do work they consider socially destructive, they consider it
a minus, but they're compensated for it. And we want them to be treated well at their work. And
that is a malleable feature of the environment. Managers can be trained to treat people better,
and they actually get better results from it. And not to be harassed, not to be the subject
of racism, not to be discriminated against in a variety of ways. But really starting,
let's start with reasonable standard of living. Let's add to that work that uses your expertise,
so allows you to be valuable in a
particular way. And also that's fulfilling people find it fulfilling to use expertise.
And then let's make sure that that is a appropriately morally positive environment.
Before I ask for the three takeaways you'd like to leave the audience with,
is there anything else you would like to mention that you haven't already touched upon?
Well, this has been a fascinating, wide-ranging conversation.
It's very easy for us to see
what's going away in terms of work.
And so whenever people see more automation,
they think, oh my God,
there's less and less for people to do.
And people have thought that
for a couple hundred years
and they've always worried about it
and not without some justification
because it really does have costs and benefits.
But there's no evidence
that we're running shy on work per se.
And moreover, we are constantly creating new work that requires new expertise, new skills,
creates variety.
Moreover, in general, I don't think work primarily exists for its own reward.
It exists to support people leading fulfilling, enjoyable, and adequately resourced lives.
And as we get more productive, we do tend
to do less work in aggregate, even though we have many jobs. People used to work 3,000 hours a year
in the US, 3,500. Now they work on average a little more than 2,000. People used to start
working when they were 10 years old, and they would keep working until the day they died.
Now people enter the labor force at 18 or 20, if you get a graduate degree at 40, and then they retire at 60 or 65
and they enjoy many years of good health while their faculties are still intact. Although we
perceive ourselves, oh, we work so hard, we work so much, we do in a condensed way. But as a fraction
of our healthy years, we work a lot less than we used to. And that's good.
That reflects our rising living standards and our success of turning rising productivity
into shared material wealth and prosperity.
David, what are the three takeaways you'd like to leave the audience with today?
One is people spend a lot of time worrying about the quantity of jobs.
We're going to run out of jobs.
There's no evidence we're running out of jobs, but they should time worrying about the quantity of jobs. We're going to run out of jobs.
There's no evidence that we're running out of jobs, but they should be worried about the quality of jobs, not the number, but do they use expertise?
Do they provide a good standard of living?
Those two things are very connected.
It's the reason we have so many jobs that are low paid is because many of them are using
generic skills.
So that's the second point actually is what matters and what has led so much to the deterioration of job quality for people
without college degrees is a lack of expert work, a lack of work that uses specific knowledge and
capacities that individuals have developed. And that could be expert work in contracting. It
could be expert work in caring for others. It could be expert work in preparing meals.
It could be expert work in caring for others in coaching.
But it matters that it's expert.
The third point I make is people spend a lot of time trying to forecast the future.
Say, oh, I can tell you what is going to happen.
If I could just project the course of this technology, I could tell you, and I know the
technology, therefore I will be able to.
But that makes a mistake about our agency. It treats the future as something that will happen
as it will happen. And our job is just to guess ahead to get the right answer. But in fact,
the future is something that we are creating. It's malleable. It depends on the choices we make,
where we invest, what we prioritize. When we think about where's the technology going,
what's AI going to do? It
really matters where we put the money. If we put it into a surveillance state, we will get a
surveillance state. AI is really good for that. But if we put it into education, we could make
education immersive, a virtual learning environment, as opposed to a classroom and cheap and accessible.
We could put it into healthcare and improve the quality and accessibility of
healthcare. Healthcare is quite expensive. Make it accessible to many more people and make the work
something that many people could accomplish with some training expertise and supported by
the technology that would support decision-making. So the future is a social creation and we make it
by setting priorities. Those priorities need to reflect
our values. So we should not ask what the future will be, but what future we want.
And then it's difficult, but we're much more likely to get that one if we know what we're aiming for.
David, this has been terrific. Thank you so much.
Thank you very much. It was a pleasure.
If you enjoyed today's episode and would like to
receive the show notes or get new fresh weekly episodes, be sure to sign up for our newsletter
at 3takeaways.com or follow us on Instagram, Twitter, and Facebook. Note that 3takeaways.com
is with the number three. Three is not spelled out. See you soon at 3takeaways.com.