00:00:00
every time we see a potentially
00:00:02
transforming
00:00:04
technology that could affect work right
00:00:06
so we see that back in the days
00:00:08
Industrial Revolution the ludites
00:00:11
movement right those people who
00:00:12
literally burn the Looms and uh it
00:00:15
didn't turns out that you know having
00:00:17
these looms that effectively accelerate
00:00:21
the process of making garments to not
00:00:23
make these people lose their jobs we
00:00:26
actually see an increase in employment
00:00:28
so every time we see technology that
00:00:31
could potentially you know change the
00:00:33
way we were change our lives we I think
00:00:35
there's like a human
00:00:37
visual reaction to it welcome to the
00:00:40
ripple effect the podcast that takes you
00:00:42
on a journey through the minds of work
00:00:44
and faculty I'm your host Dan Loney and
00:00:47
in each episode we'll be diving deep
00:00:48
into the inspiration behind the
00:00:50
groundbreaking research that Wharton
00:00:52
professors have conducted and exploring
00:00:55
how their findings resonate with the
00:00:57
world today we'll be covering a diverse
00:00:59
range of topics bringing you the latest
00:01:01
insights and knowledge that you can
00:01:03
apply to your life into work so get
00:01:06
ready to dive into new ideas with the
00:01:09
ripple effect
00:01:10
and a pleasure to be joined by Lin Wu
00:01:12
who's associate professor of operations
00:01:14
information and decisions here at the
00:01:16
Wharton School Lynn great to talk to you
00:01:18
again
00:01:20
so much for having me thank you and I
00:01:22
guess let's start out by what it was
00:01:24
that has kind of originally sparked your
00:01:27
interest in studying how kind of robots
00:01:30
are are starting to change employment
00:01:32
that's a great question honestly this
00:01:34
one has the easiest motivation at the
00:01:36
time they're I mean still going on now
00:01:38
there were just so many articles on a
00:01:41
press and Academia or everywhere about
00:01:43
how robots are going to take over all
00:01:45
our jobs and the pending robocalypse
00:01:48
that's gonna happen very soon and uh
00:01:51
there were already policy pieces on
00:01:54
robot taxes you know you had one Bill
00:01:57
Gates promoted it Bill de Blasio made it
00:02:00
a central piece of the presidential
00:02:02
campaign back in 2000 2020. and the
00:02:05
Bernie sander recently
00:02:07
had also proposed some kind of a robot
00:02:09
attacks so I think it's really important
00:02:12
to understand
00:02:14
what's going on here like have some real
00:02:16
concrete evidence
00:02:18
at the firm level to see whether firm
00:02:21
actually do laid off people on mass
00:02:23
after robot adoption and there were only
00:02:26
industry in the country level evidence
00:02:29
at the time and there's actually really
00:02:31
important to study at the firm level
00:02:33
because you know countries and
00:02:35
industries do not adopt robots firms do
00:02:37
right and uh whatever the positive
00:02:40
negative fast you find depends on you
00:02:43
know whether firms that adopt actually
00:02:46
are people were they you know where is
00:02:48
actually coming from from that do not
00:02:49
adopt so this kind of a you know this
00:02:51
kind of I found a very difficult to
00:02:53
observe in when you're looking at a
00:02:54
macro level at industry and Country
00:02:56
level why is it you think we've kind of
00:02:58
had these narratives pop up and and and
00:03:01
really in many cases uh they have taken
00:03:05
hold uh in with some people over the
00:03:07
last few years
00:03:09
I think it's uh not just last few years
00:03:12
like I think I've seen this kind of
00:03:13
Trends going on like every time we see a
00:03:17
potentially transforming
00:03:19
technology that could affect work right
00:03:22
so we see that back in the days
00:03:24
Industrial Revolution the ludites
00:03:26
movement right you know there's people
00:03:27
who literally burn the Looms that
00:03:30
actually automated the process of making
00:03:32
uh you know garments
00:03:35
and uh it didn't turns out that you know
00:03:38
having these uh looms that effectively
00:03:41
accelerate the process of making
00:03:43
garments did not make these people lose
00:03:46
their jobs we actually see an increase
00:03:48
in employment
00:03:49
for people who can effectively use these
00:03:52
new automated mechanical tools like like
00:03:55
looms and you know we see the same thing
00:03:57
with like you know Excel excels like
00:03:59
it's going to take it's going to replace
00:04:00
accountants it never happened right so
00:04:03
every time we see technology that could
00:04:06
potentially you know change the way we
00:04:08
work change our lives we I think there's
00:04:09
like a human
00:04:11
visual reaction to it and especially we
00:04:14
tend to overestimate what a technology
00:04:16
technology can do
00:04:17
and thinking oh my gosh I'm gonna lose
00:04:20
my jobs now and I think it's really
00:04:22
important to take bad to think about
00:04:23
what exactly is that technology doing
00:04:25
the four
00:04:27
you know we make any strong decisions
00:04:30
especially the policy makers uh you know
00:04:32
a policy angle to decide you know what
00:04:35
are we going to do about this what are
00:04:37
workers going to do about what firms
00:04:39
they're going to do about it
00:04:40
so tell us a little bit about the
00:04:41
research that you're doing in this area
00:04:43
to try and really get a better grasp on
00:04:45
what's what's been taking place here
00:04:48
yeah absolutely so our work is a first
00:04:52
to study robot adoption unemployment how
00:04:55
the facts are going is going to be at a
00:04:57
firm level so one emphasis at the firm
00:04:59
level is because
00:05:01
only at a firm you can see what happens
00:05:03
when firm adopt robot do they actually
00:05:05
lay off people or do they actually hire
00:05:08
more people and what happened to The
00:05:10
Firm that did not adopt right so these
00:05:12
kind of uh effects can only be observed
00:05:14
at a more micro level
00:05:16
so we actually use the data from
00:05:18
statistical Canada which has a
00:05:21
comprehensive data on robot Import and
00:05:24
Export so which have a very good measure
00:05:26
of robot adoption data and we also have
00:05:30
very comprehensive data about the
00:05:32
financial performance what about from
00:05:33
the tax filings a bunch of uh you know
00:05:36
mandatory surveys that the Canadian
00:05:39
government mended it
00:05:40
various firm practices
00:05:43
so we found is exactly some opposite of
00:05:46
what people were expecting that robot
00:05:49
did not replace human workers in fact
00:05:53
the robot adapters or the firm that
00:05:55
adopted robots hired more people than it
00:05:59
did before
00:06:01
so how do we reconcile the evidence that
00:06:03
we see sometimes an industry level the
00:06:06
country level there is a negative effect
00:06:08
on robots unemployment
00:06:11
it turns out it's not a robot adopting
00:06:14
firms that are hurting employment
00:06:16
it is a firm that did not adopt robots
00:06:20
that are losing the competition
00:06:22
they're not competitive as before and
00:06:24
they had to lay off people because
00:06:26
they're losing the market share
00:06:28
so it's a very different
00:06:30
story than the popular press thinking
00:06:33
that we got a tax robot to preserve
00:06:34
human work it turns out is the People
00:06:38
The Firm that did not adopt robots need
00:06:40
help so taxing the firm that robots or
00:06:44
turned out to be exactly the wrong thing
00:06:45
to do in this case
00:06:48
that's so that's like a one major
00:06:49
finding in in in saying in basically
00:06:53
saying well look we actually look at
00:06:54
these phenomena in a greater detail to
00:06:57
understand what's going on here without
00:06:58
this kind of a firm level measurement
00:07:01
and and Technology measurement we
00:07:03
wouldn't be able to know this important
00:07:05
distinction
00:07:07
we also have found other effects on
00:07:10
employment it's not the number that
00:07:12
matters so we always think that robots
00:07:14
are you know taking over our jobs and
00:07:17
that's not that's not the case but it
00:07:19
turns out that the story is not as rosy
00:07:23
as Swig had expected right so we
00:07:26
actually see the skill effects basically
00:07:30
the federal robot on different type of
00:07:32
skills
00:07:33
has a very different has a different
00:07:35
story so specifically
00:07:37
robot adopting firms hired more High
00:07:40
skill workers
00:07:42
and many more low skill workers at the
00:07:46
expense of middle skilled workers so
00:07:49
here I Define High skill worker are
00:07:51
those with college education low skill
00:07:53
workers are the people with barely
00:07:55
finished high school and middle school
00:07:58
workers are people with high school
00:07:59
degrees but with kind of where associate
00:08:01
degrees has some kind of a more advanced
00:08:04
to work related trainings
00:08:06
so is this middle skill workers are
00:08:08
being decimated
00:08:10
by robot and that is a big problem and
00:08:13
we also show that not only that
00:08:15
managerial work supervisory work has
00:08:18
also been decimated by robots
00:08:21
so
00:08:22
if you look at average number it looks
00:08:25
great like employment has gone up but by
00:08:27
hollowing all the middle skill work
00:08:29
hollowing out the supervisory work is
00:08:30
actually a big problem because now the
00:08:33
career ladder is broken right right so
00:08:35
how do we incentivize how do we train
00:08:37
how do we make sure uh you know where
00:08:40
did middle skill work no they can't I
00:08:42
mean you can't expect people to all get
00:08:43
college degrees and become you know
00:08:44
programmers or become robot technicians
00:08:46
right or you know producers right
00:08:50
yeah same thing with old skills yeah
00:08:53
does the adoption of robots in the firms
00:08:57
that you see uh that have done that does
00:09:00
it change the Dynamics of the work being
00:09:02
done by those companies not even
00:09:05
necessarily the labor side of it but
00:09:07
actually the the actual work and maybe
00:09:09
even the success rate for the company
00:09:12
itself
00:09:13
absolutely so let me give you an example
00:09:15
right so I think you touch on a very
00:09:18
important question is what like you know
00:09:19
it's just adopting the robot itself is
00:09:21
not going to be enough right you have to
00:09:24
be able to learn how to manage robots in
00:09:28
a way that accelerates your performance
00:09:31
it increase your work or productivity so
00:09:33
let me give you an example a real live
00:09:35
example it's like a repair facility at a
00:09:37
U.S electronic firm they actually
00:09:40
experience a dramatic Improvement
00:09:42
in their ability to observe productivity
00:09:46
after robots were implemented
00:09:49
okay so this is a repair facility they
00:09:51
fix Electronics okay and because robots
00:09:54
don't get tired they don't get
00:09:55
physically tired when performing this
00:09:57
kind of repetitive task of you know of
00:09:59
fixing certain uh errors a certain
00:10:02
problem in the electron electronic board
00:10:05
so they can do this job more
00:10:07
consistently than humans who previously
00:10:10
doing the same task
00:10:11
as a result of variance in production
00:10:13
actually have gone down so this allowed
00:10:16
manager to clearly observe individual
00:10:18
employees behaviors okay and they
00:10:21
actually found that you know through
00:10:22
this system because robots are doing
00:10:24
these kind of tasks they were able to
00:10:25
see oh many employees so human employees
00:10:28
were following irregular patterns or
00:10:29
being very productive in the morning
00:10:30
compared to the afternoon
00:10:32
and in the afternoon their productivity
00:10:35
kind of went down a little bit
00:10:37
and but then they do more repairs later
00:10:39
hours as they were like you know
00:10:41
cramming their work End of Days
00:10:43
so interestingly after robots are
00:10:46
implementing the repairing process right
00:10:48
they are actually able to trap as
00:10:50
individual employees productivity for
00:10:53
easily or for two reasons right first
00:10:55
the type of Errors the robots made are
00:10:58
very different from that of humans okay
00:11:00
so because of the differentiating errors
00:11:03
between humans and the robots is make it
00:11:06
easier for us to figure out which one is
00:11:07
which
00:11:08
okay and robots are also more likely to
00:11:10
make consistent errors compared to human
00:11:12
errors right again making human errors
00:11:15
easier to identify okay another reason
00:11:17
is that robots also provide a precise
00:11:21
data about their own performance okay
00:11:22
which also made it easier to isolate
00:11:25
both positive and negative side of
00:11:27
performance changes caused by human
00:11:29
behaviors so this data generating
00:11:31
capability allow manager to you know
00:11:34
monitor their productivity much better
00:11:35
than before detect weaknesses in the
00:11:37
production processes and then still just
00:11:40
not just stopping robot itself all the
00:11:42
other thing they've done to detect the
00:11:44
figure out in the production production
00:11:45
process
00:11:46
in this case the manager of the referred
00:11:49
facility wasn't even aware of this
00:11:51
cramming Behavior described earlier
00:11:53
until the robots were able to Monitor
00:11:55
and adopt uh other than that robot to
00:11:58
observe these processes
00:12:00
and as a result they were changing lives
00:12:02
of the word processes along the way to
00:12:05
further reduce the errors once human or
00:12:08
human workers and robots are working
00:12:10
together and overall error raised in a
00:12:12
facility has dramatically improved
00:12:13
that's an example of where robots how
00:12:16
robots can be used to improve
00:12:18
productivity in human uh how to how to
00:12:20
manage Workforce appropriately to
00:12:23
capture to further increase the effect
00:12:25
of the robots
00:12:26
how do you think that then careers are
00:12:30
going to be impacted by the further
00:12:33
adoption of robots and and I'll play
00:12:34
that off of uh the comment you made
00:12:37
earlier about the impact being uh
00:12:39
significant on middle managers as we go
00:12:41
forward you would think that's a that's
00:12:43
an important kind of stepping stone in a
00:12:46
person's professional career
00:12:48
if we have fewer of those then that
00:12:51
changes I think the the hierarchy the
00:12:53
structure of leadership within companies
00:12:56
that's a really really important point
00:12:58
and it's actually a really hard problem
00:13:00
to solve so in my research I mentioned
00:13:02
earlier because you have many more lower
00:13:06
skill workers many more higher skill
00:13:08
workers at expensive Metals your work
00:13:10
the type of manager you need is going to
00:13:14
be very different right these managers
00:13:16
need to understand how a robot Works to
00:13:18
be able to detail these processing just
00:13:20
just like just like the example I gave
00:13:22
about repair facilities right they need
00:13:24
to change fundamentally change the way
00:13:26
they work fundamentally change the world
00:13:28
the way they monitor and reward
00:13:30
and hire employees right so these are
00:13:33
those are two effects number one we
00:13:35
simply need fewer managers than before
00:13:37
supervisors before because managing
00:13:39
standardized work on Lower skilled
00:13:40
worker is you can manage many at the
00:13:43
same time as opposed to higher skill or
00:13:45
middle skill workers right and
00:13:47
furthermore the type of manager skill
00:13:49
management skills you need is going to
00:13:50
be different right so it's a big problem
00:13:54
right so now I so now we have no middle
00:13:57
skill work but less of them and much
00:14:00
fewer supervisory work and where do
00:14:02
people go right so middle so the entry
00:14:05
level work is supposed to be a stepping
00:14:06
stone as you said to move up in the
00:14:09
career hierarchy
00:14:10
and now you can't move up anymore right
00:14:12
you've got middle still you're gone you
00:14:14
go to uh the supervisors and you are
00:14:17
gone
00:14:19
and then you can't it's very hard to
00:14:21
move up to high skill work right that
00:14:22
requires extensive training so this is a
00:14:25
very big challenge for managers right
00:14:27
you had to think about well how do I
00:14:29
build a new career ladder for my
00:14:32
employees now the existing one may not
00:14:35
work it may not have worked already
00:14:39
and that's why you see a lot of
00:14:42
unionization going on in the workforce
00:14:45
from Amazon warehouse to Starbucks
00:14:47
everywhere it's because that you cannot
00:14:49
you can't use a career ladder as a
00:14:51
motivating force for people to you know
00:14:55
work in the entry-level job by lower pay
00:14:57
in in exchange for a future career
00:14:59
advancements
00:15:01
so how do you build that back you know
00:15:04
if you don't the firms do not build a
00:15:05
back then we're going to see Union
00:15:07
becoming uh you know a more Mainstay in
00:15:10
our in our society again so if we're
00:15:12
expecting that we're going to see more
00:15:14
companies uh adding robotics uh into
00:15:17
their operational structure here in the
00:15:19
years ahead I guess does it answer the
00:15:22
question whether or not companies can
00:15:24
even avoid having robots in the first
00:15:26
place it almost seems a little bit like
00:15:28
they can't afford to do that as we move
00:15:31
farther down the road yeah this this
00:15:34
case unless there's no one adopting a
00:15:37
technology the moment one one firm
00:15:39
adopts they become more effective more
00:15:40
competitive that means everyone else in
00:15:43
order to stay competitive in the market
00:15:45
in Marketplace you have to adopt these
00:15:48
new technologies right like you just uh
00:15:50
even the biggest firms like you know um
00:15:52
very profitable very you know you know
00:15:56
very you know Innovative firms have to
00:15:59
catch up on that game and we've seen
00:16:01
that Google's case when Microsoft
00:16:03
released our GPT Google's scrambling to
00:16:06
do the same thing you know incorporating
00:16:09
every aspect of that technology in their
00:16:10
in their products like that is something
00:16:12
that firms cannot avoid uh you know just
00:16:15
uh burning the ludice right the little
00:16:18
guy's burning the Looms not gonna work
00:16:19
yeah
00:16:22
how then do you think that uh the
00:16:25
advancement of chat GPT will play a role
00:16:27
uh in the corporate uh corporate
00:16:30
structure as we move forward
00:16:33
I think
00:16:35
uh for the
00:16:37
LGBT large and large language models is
00:16:39
going to accelerate that process
00:16:41
tremendously because you think about
00:16:43
What Chai GPT and the this large
00:16:46
language models you know what they're
00:16:48
targeting is exactly that middle skill
00:16:50
work
00:16:51
right it's not like it's entry level
00:16:53
work like you feel like you know and uh
00:16:56
that that will be replaced and then
00:16:59
because these technology are really good
00:17:01
when you already know when you are an
00:17:03
expert in that field already
00:17:05
accelerates your work but who are doing
00:17:07
those work for you before if you were
00:17:09
you know
00:17:10
a senior person is people below you now
00:17:13
you can use the charge EBT to do a lot
00:17:15
of it for you so it's precisely that
00:17:17
middle skill work that is being targeted
00:17:19
and that's exactly you know it's the
00:17:21
same effects of robots and there was
00:17:23
like a recent paper by openai and also a
00:17:26
colleagues at Wharton I showed that the
00:17:28
the the the skill set up being targeted
00:17:30
gbt is programming drops again middle
00:17:33
skill programming jobs and writers
00:17:35
writing jobs again those are you know
00:17:38
entry-level writing jobs so those are
00:17:40
being you know being massively
00:17:42
accelerated by the chat NDP so is the is
00:17:46
the middle skilled job uh does its
00:17:50
longer term uh outlook look very Bleak
00:17:54
at this point or is there going to be
00:17:57
some level of middle skill work that
00:17:59
will still be there in conjunction with
00:18:01
all of this advancement around
00:18:03
technology
00:18:05
see middle skill work is in trouble
00:18:08
but new middle skill work will be
00:18:10
created
00:18:11
for example right prompt engineering
00:18:13
something you've never heard of until
00:18:15
maybe it's a few months ago these
00:18:17
Engineers are literally trying to make
00:18:19
happy do what it's supposed to do
00:18:22
right prompt engineering or robot
00:18:24
technicians to fix the robots right
00:18:26
right process engineering to think
00:18:28
observe processes to see where robots
00:18:31
can be used in the production processes
00:18:33
right all these things are you know
00:18:35
probably going to be new tasks and then
00:18:38
over time they'll tell new tasks will be
00:18:39
involved into new career opportunities
00:18:42
just like you know 20 years ago there
00:18:44
was no social media manager right that's
00:18:46
a new job that created as a result of
00:18:49
the Technologies but the important
00:18:52
problem is not necessarily the new job
00:18:53
will be created I guarantee you that new
00:18:56
jobs and new tasks will be created is
00:18:59
the speed at which we can retrain the
00:19:01
existing Workforce to actually leverage
00:19:03
that right the last time we had this
00:19:06
kind of dramatic technology change is
00:19:09
probably Industrial Revolution
00:19:11
um steam engine replay being replaced by
00:19:13
electricity that took 30 40 years to
00:19:16
complete right and that means a new
00:19:18
generation of managers retired New
00:19:20
Generation sorry the existing like uh
00:19:22
you know existing managers to retire
00:19:24
existing Workforce retire for new
00:19:26
managers a new Workforce who are already
00:19:28
you know being trained in the time a lot
00:19:30
of Technology too more effectively
00:19:33
except in this time
00:19:35
we're not going to have 30 40 years
00:19:37
Horizon we're gonna have five ten years
00:19:39
Horizon so how do you retrain that
00:19:43
existing Workforce
00:19:44
is going to be a huge challenge for
00:19:46
everyone for firms for policy makers for
00:19:49
everyone yeah
00:19:51
thank you for listening to the ripple
00:19:52
effect we hope you found this episode
00:19:54
informative and engaging don't forget to
00:19:56
subscribe and leave us a review so that
00:19:59
we can continue to bring you the best
00:20:01
Insight from the Wharton School
