00:00:00
well artificial intelligence has been
00:00:01
around for a while but it seems like
00:00:04
2023 was the year that it took over as
00:00:06
our primary focus when thinking about
00:00:09
the future of business and other aspects
00:00:11
of Our Lives Christian tursh is a
00:00:14
co-director of the Mac Institute for
00:00:16
Innovation management he is also
00:00:18
professor of operations information and
00:00:20
decisions here at the Wharton School and
00:00:22
he joins us here in studio great to see
00:00:23
you thanks for having me here how do you
00:00:26
process all that we have seen and talked
00:00:29
about around AI this year it's been a
00:00:32
crazy year so if you think back to the
00:00:34
fall of
00:00:35
2022 when chbt came out it started off
00:00:39
as something that we thought of
00:00:40
primarily the kids would use to get
00:00:42
their homework done and it took the
00:00:44
World by storm at the same time we had
00:00:48
Dar mid Journey stable diffusion do
00:00:50
amazing work in in the texture image
00:00:53
world and with GPT 4 launch the the
00:00:56
large language model has got even better
00:00:58
it's been a hell of a ride I would
00:01:00
imagine it's probably not even a
00:01:02
surprise to you to see how it has kind
00:01:05
of enveloped the thinking of so many
00:01:07
businesses over the last year as well so
00:01:10
I
00:01:11
think we had a big jump in progress I
00:01:14
think technological progress doesn't
00:01:16
typically follow Continuum there these
00:01:18
what we call punctuated equilibria these
00:01:21
moments when there are big shifts and I
00:01:23
think we have seen this massive shift
00:01:25
that go beyond what companies expected
00:01:28
and I lots of companies work with at the
00:01:30
mech Institute are now thinking about
00:01:33
getting ready for the changes and
00:01:35
retooling so to say now that the new
00:01:37
tool is available is this going to be uh
00:01:41
a a path that all businesses will be
00:01:43
able to take advantage of I think the
00:01:46
expectation is obviously larger
00:01:48
companies that have the resources will
00:01:49
easily be able to do this but I guess my
00:01:52
question is more about what about the
00:01:53
midsize and small businesses and how
00:01:56
they may be impacted by this so that
00:01:58
they're not necessarily left out I think
00:02:00
it's actually the opposite I think this
00:02:03
change that we have seen is going to
00:02:05
lead to a democratization of AI it used
00:02:07
to be that you have to be a big company
00:02:09
to have the the big data the the huge
00:02:12
data set that you use to train your AI
00:02:14
yourself and I think thanks to the GPT
00:02:17
technology basically their training is
00:02:19
it's pre-trained now and so right out of
00:02:22
a box uh a one person startup can create
00:02:26
a kind of a cool app uh and provide
00:02:29
solution to needs that are currently
00:02:30
unmet so I think it helps the small guys
00:02:34
the one thing that we've seen with the
00:02:36
updates of uh C GPT and open Ice workers
00:02:40
of course that every time that they come
00:02:42
out with a new release thousands of
00:02:44
these small Ventures that were about to
00:02:46
take off are wiped away because their
00:02:47
functionality is faked into the Next
00:02:49
Generation but I I don't think it is
00:02:52
giving the big guys an unfair Advantage
00:02:56
it feels almost a little bit like we're
00:02:58
in a process right now where it's like
00:03:00
when we get an update for our smartphone
00:03:02
or you know have to update our our uh
00:03:05
our our security uh on our computer
00:03:07
right now that it's coming at that speed
00:03:09
almost unbelievable right you go from
00:03:11
3.5 to four to Turbo the gets integrated
00:03:15
I mean all in one year after again there
00:03:18
been progress in AI for for for many
00:03:21
decades but uh the last year it seems to
00:03:25
have hit the mainstream both on the
00:03:28
availability of the techn
00:03:30
technological solution capabilities but
00:03:33
also now I think the awareness of what
00:03:34
it can do in Enterprises where we can
00:03:36
use it to create value you and I had
00:03:39
talked uh many months ago when you had
00:03:41
done research around uh the impact of AI
00:03:45
on uh it trying to pass an NBA test uh
00:03:51
how has your thinking around the use of
00:03:54
AI in the education setting developed
00:03:57
over that time as well so the first
00:03:59
thing I think most of us do now is if we
00:04:01
have an exam that is kind of a
00:04:03
traditional exam uh B GP B GPT right I
00:04:07
mean it's too good it passes every exam
00:04:10
you can't figure that out or detect that
00:04:14
at the end of the day we at the world in
00:04:15
school want to give degrees to students
00:04:18
and not to to software having said this
00:04:21
uh I have in my operation scores just
00:04:24
kind of created the new first new
00:04:26
assignment where I explicitly asked
00:04:28
students to use G PPT to help them sort
00:04:31
through massive and complex data uh I
00:04:34
have in my Innovation course where we
00:04:35
start new Ventures and come up with kind
00:04:37
of ideas on new products and services I
00:04:41
urge my students to use CET GPT to
00:04:43
generate ideas and even in something as
00:04:46
bin is looking at my teaching
00:04:47
evaluations where we get lots of verbal
00:04:49
comments I just put this into GPT and
00:04:52
say like summarize this and make three
00:04:54
recommendations of what I should improve
00:04:55
next year and it's pretty good but so
00:04:59
you don't have a negative viewpoint on
00:05:01
it at all it's very much a positive you
00:05:04
know and the benefits he can provide
00:05:06
Viewpoint well maybe that is wishful
00:05:08
thinking but I mean we don't have a
00:05:10
choice right but I think look
00:05:13
if all we did is we would go back to a
00:05:17
world where we did not have this
00:05:19
technology uh it would be a waste right
00:05:21
this is a huge opportunity for many
00:05:24
professions Education Health Care uh and
00:05:27
many others of really dealing with the
00:05:30
big problems I mean a lot
00:05:32
of jobs uh folks are burning out folks
00:05:37
are working I mean we had we had a
00:05:38
meeting with American Institute our
00:05:39
board last night folks from pen medicine
00:05:42
were there uh the burnout rate of
00:05:45
providers of people working in medicine
00:05:47
is huge we need to help these people and
00:05:50
so if you if you give me a a magic want
00:05:53
that makes me a 50% productivity
00:05:55
Improvement I take the
00:05:57
want so you just talked about about
00:06:00
impact uh one of the areas of impact
00:06:04
that has been brought up is is AI going
00:06:08
to eventually take jobs away but there's
00:06:11
also been a focus on how AI can support
00:06:14
the efforts of a lot of people uh in the
00:06:17
workplace absolutely right which way do
00:06:19
you think it's going to go no I think
00:06:21
both right I mean so there will be jobs
00:06:24
that will change I mean basically any
00:06:26
job to be done there is a customer who
00:06:28
needs something
00:06:29
there is a provider that provides a
00:06:32
solution and they have to do this in a
00:06:34
way that value is created so that means
00:06:36
the customer is willing to pay more than
00:06:37
it costs to generate solution so now
00:06:40
first order effect is you have ai
00:06:42
generates a solution which is much
00:06:44
cheaper than the human generated
00:06:45
solution and so lots of people will have
00:06:48
to change their jobs but second order
00:06:51
there are lots of things that right now
00:06:52
make no economic sense of doing unmet
00:06:55
customer needs that we with AI can
00:06:58
fulfill which which means there're new
00:07:00
jobs and which way that goes I don't
00:07:02
think anybody knows I I know you've also
00:07:05
taken a look at at AI in the scope of
00:07:08
entrepreneurism as well how does that
00:07:12
impact being an entrepreneur moving
00:07:13
forward so I think one of the great uses
00:07:17
of AI is as a form of a brainstorming
00:07:22
partner we we still see hallucinations
00:07:25
in AI we don't think we don't have the
00:07:27
the trust in the technology yet where we
00:07:30
would say like drive my car or uh manage
00:07:34
my retirement Port portfolio the nice
00:07:36
thing was entrepreneurship in idea
00:07:38
generation is you're giving me a 100
00:07:40
ideas even if 99 of those 100 ideas are
00:07:43
horrible as long as there's one good one
00:07:45
in the middle I I take that portfolio
00:07:48
and so I think that startup aspect of
00:07:51
generating ideas for dissertations for
00:07:54
cancer research for
00:07:56
entrepreneurship that part of creativity
00:07:58
we can can enhance now and that's that
00:08:02
to me is a no-brainer I take that any
00:08:04
time you bring up something interesting
00:08:05
as well because uh there are still calls
00:08:09
to maybe slow down the Implement
00:08:11
implementation of AI so that we truly
00:08:14
understand what the impact is where do
00:08:17
you fall on that on that
00:08:18
question I mean it's uh probably In Our
00:08:21
Lifetime the biggest game changer that
00:08:24
we've seen in history I think you and I
00:08:26
have been around for the internet
00:08:27
becoming a big thing right then mobile
00:08:30
came I think this is probably the
00:08:31
biggest thing in our lifetime uh so it's
00:08:36
really hard to predict right I mean you
00:08:37
see people making the argument for super
00:08:41
intelligence and that the robots will
00:08:43
rule the world smart people argue so uh
00:08:46
I think there are lots of things we
00:08:48
regulate in our society that is a good
00:08:50
thing uh you're you're not not allowed
00:08:52
to build a nuclear bump in your backyard
00:08:55
and again that's a good thing so it
00:08:57
would be shocking if we would leave this
00:08:59
space is totally unregulated I think
00:09:01
what is especially concerning here is
00:09:04
that if you're building an atomic bomb
00:09:06
it took the Manhattan Project over
00:09:08
100,000 people to build an atomic bomb
00:09:10
right I mean that's something that you
00:09:12
and I even with evil intent couldn't do
00:09:14
in our
00:09:15
backyard uh you jailbreak a a smart AI
00:09:19
system it could do horrible thing it
00:09:22
would only take one offender and so I
00:09:24
think some form of security some form of
00:09:26
Regulation absolutely is your
00:09:28
expectation then that the next couple of
00:09:31
years we'll see even greater impact from
00:09:34
from things like chat GPT moving
00:09:37
forward I mean the first order of
00:09:39
forecasting is extrapolation right I
00:09:41
mean you basically extrapolate the line
00:09:43
and uh that's suggest that we're going
00:09:46
to see ma massive progress It's not
00:09:48
going to stop tomorrow right so we're
00:09:49
going to see massive progress how long
00:09:51
this famous discussion of AGI super
00:09:53
intelligence how far we are away uh
00:09:56
nobody knows C certainly I don't know
00:09:59
but I think that is more Phil more of a
00:10:01
philosophical but it it still feels like
00:10:05
from all that's happened and and with
00:10:07
the potential impact of this technology
00:10:10
now you know in our Labs that it feels
00:10:13
like it's still a little bit of the tip
00:10:14
of the iceberg doesn't
00:10:16
it yes and I think again it will not
00:10:20
stop if you even for just another one or
00:10:23
two years you apply some form of
00:10:25
extrapolation of the trend that we've
00:10:27
seen over the last years uh we're going
00:10:29
to have jaky rolling level writing come
00:10:33
out of CH GPT we could imagine the
00:10:36
context window has just grown
00:10:38
substantially we could imagine much
00:10:41
better science papers and dissertations
00:10:44
be generated by CH GPT I don't think it
00:10:47
takes much of a crazy assumption to
00:10:51
imagine a world in four or five years
00:10:54
where a lot of these things become
00:10:56
true from a business perspective then
00:10:59
one of the things I wondered is does
00:11:02
this drive even a greater focus on what
00:11:05
productivity will be in the future with
00:11:08
employees and getting projects done
00:11:11
because you will have this component as
00:11:12
an assistant to be able to get a lot of
00:11:14
this work uh done in a quicker fashion
00:11:18
but also potentially in a in a more
00:11:20
concise fashion as well I'm a big fan of
00:11:22
productivity it sounds totally boring
00:11:26
German operations management type of
00:11:27
boring uh but again think about
00:11:31
professions such as healthc care think
00:11:33
about
00:11:34
Educators we need to make them more
00:11:37
productive to improve their quality of
00:11:39
life and have healthier patients and
00:11:41
smarter children right I think those
00:11:44
productivities uh it's up for us as a
00:11:47
society what we do with these
00:11:48
productivity gains and I very much hope
00:11:51
that we use them for better education
00:11:54
and better treatments as opposed to for
00:11:56
lower cost of healthare and
00:11:59
so you mentioned earlier about the
00:12:01
element of stress that a lot of
00:12:03
companies are dealing with in healthcare
00:12:05
as well are there other avenues that
00:12:08
that the hospital industry is watching
00:12:11
closely where they think that AI will be
00:12:13
able to to provide a benefit moving
00:12:15
forward I think in the news
00:12:19
the stories that that reported is the
00:12:23
capabilities of AI to help with the
00:12:24
diagnosis right it it can read uh in
00:12:28
image to text it can read x-rays it can
00:12:32
uh I'm working on a project on Mental
00:12:34
Health right now where it can do pretty
00:12:36
decent diagnostic
00:12:38
assessments uh I think we forget how
00:12:40
much of the of of the life in a a day of
00:12:44
the life in the worker in healthcare
00:12:46
nurse or provider how much that is
00:12:49
writing notes reports keeping track of
00:12:52
medical records I think a lot of that
00:12:54
can be freed up and I think that's a
00:12:56
much low hanging frood it's less
00:12:58
glorious it's l sexy to talk about yeah
00:13:00
but I think that's where the first line
00:13:01
of attack is going to be but I would
00:13:03
imagine that it's partially benefiting
00:13:05
in in an area like healthcare from the
00:13:08
fact that the healthcare industry has
00:13:09
already kind of made a transition to
00:13:11
technology to begin with and so this is
00:13:13
the next step in the process absolutely
00:13:15
I I come from Germany my parents live in
00:13:18
Germany that care in Germany
00:13:21
unbelievable how this system is still
00:13:23
paper based and years behind we have had
00:13:27
lots of people joke complain about the
00:13:29
transition to the electronic health
00:13:31
record it's been uh the right thing and
00:13:34
I think we're going to see massive
00:13:35
rewards from that what are you most
00:13:38
interested to watch or potentially see
00:13:41
occur next year in and around this
00:13:44
area I think next year I want to see how
00:13:48
we speaking for myself how we go from
00:13:51
cool projects of guess what it can do
00:13:54
that towards really uh
00:13:57
implementing uh at the bedside or at the
00:14:01
school bench implementing systems that
00:14:04
lead to better outcome I think we we
00:14:06
urgently needed and this is going to be
00:14:08
that next phase of learning for so many
00:14:10
of the younger Generations is how AI can
00:14:14
complement a lot of what we do as we you
00:14:17
know as we are growing up and going out
00:14:18
into the world yeah and I mean many
00:14:20
years ago you and I had discussions
00:14:22
about moves right massive open online
00:14:24
sources I mean we were 10 years younger
00:14:27
forgot about that yes and so there we
00:14:29
were saying how the moves will disrupt
00:14:31
everything but what we couldn't do with
00:14:33
a video Based Teaching is it
00:14:35
personalized tutoring and that I think
00:14:38
is a major strength that we have the
00:14:40
technology now every kid can be picked
00:14:43
up where he or she stands and be coached
00:14:47
and again I'm in no way implying that we
00:14:49
don't you need human workers teachers to
00:14:52
do this right but we can customize the
00:14:54
education to an extent that we could not
00:14:57
do this before and I think I very much
00:14:59
hope that this overcomes a lot of the
00:15:01
quality discrepancies around schooling
00:15:03
the oecd just this week released a pizza
00:15:06
reports with horrible report cards on
00:15:09
math and science for many of the
00:15:11
developed Nations kids have not
00:15:13
recovered from covid in the classroom
00:15:15
the the test results are horrible and so
00:15:18
again we need that technology we need
00:15:20
that productivity boost and shouldn't be
00:15:23
scared of it only by putting it into
00:15:26
this kind of it's going to steal our
00:15:27
jobs Christian great to talk with you
00:15:29
thanks very much always a pleasure thank
00:15:31
you Christian tursh who is uh
00:15:33
co-director of the Mac Institute for
00:15:35
Innovation management and also professor
00:15:37
of operations information and decisions
00:15:39
here at the Wharton
00:15:45
School
