00:00:01
this podcast is brought to you by
00:00:03
knowled of
00:00:04
[Music]
00:00:12
warten the Wharton School is leveling up
00:00:14
its teaching and its research in the
00:00:16
areas of artificial intelligence and
00:00:18
data science the school has launched the
00:00:20
warden Ai and analytics initiative it's
00:00:23
an interdisiplinary collaborative effort
00:00:25
to shape how generative AI will be used
00:00:28
for business Innovation and Beyond
00:00:30
welcome to knowledge at Wharton I'm
00:00:32
Angie basuni with me today is Eric
00:00:34
bradow he's a longtime faculty member in
00:00:36
our marketing department and he's an
00:00:37
applied statistician and many of you
00:00:40
probably know him from his Moneyball
00:00:42
podcast on
00:00:43
SiriusXM Eric thanks for being with me
00:00:46
oh it's great to be here Angie and
00:00:47
always happy to talk about our new
00:00:49
acronym w a AI oh goodness well you've
00:00:53
been named as the new Vice Dean for this
00:00:55
initiative so you're giving us a little
00:00:56
bit of time today to pick your brain
00:00:58
about it I want to start with ask asking
00:01:00
you to explain just a quick overview of
00:01:02
the initiative and why the school is
00:01:04
launching it right now yeah I mean it's
00:01:06
really as our Dean Erica James says it's
00:01:08
really one of those meet the- moment
00:01:10
types of opportunities for the Wharton
00:01:11
School um actually we've been in the
00:01:14
analytics business if you'd like for 20
00:01:16
years uh a colleague of mine Pete fader
00:01:19
and I started the first ever data
00:01:22
science center if you'd like at any
00:01:24
business school in the world 20 years
00:01:25
ago now and so we always believe that
00:01:29
data science basically using empiricism
00:01:32
using high-powered statistical models
00:01:34
was going to be a big part of the future
00:01:36
of business because of technology and
00:01:38
new data streams that emerge but what's
00:01:40
happened of course in the last almost
00:01:42
two years now is generative AI now
00:01:44
fortunately at Wharton we've had a
00:01:46
center called AI at Wharton for almost
00:01:48
seven years now we knew this future was
00:01:50
coming we didn't know how fast and how
00:01:52
if you like ubiquitous it was going to
00:01:54
be but the goals of our Center are
00:01:56
actually quite simple there are four
00:01:58
major goals first we want to be thought
00:02:01
of as the number one knowledge creators
00:02:04
of Business Schools in the world number
00:02:06
two we want to impact students now
00:02:09
broadly defined not just students at the
00:02:11
Wharton School or even the University of
00:02:13
Pennsylvania but we want to democratize
00:02:16
Ai and business education throughout the
00:02:18
world third we want to have an impact on
00:02:20
the SE Suite we're a business school and
00:02:22
so what I always say is we're training
00:02:24
people to be the leaders in Ai and data
00:02:27
science but not data scientists we're
00:02:29
training people ble to lead companies
00:02:31
through Ai and data science and last but
00:02:34
not least I've always called it
00:02:35
analytics for good but how about we do
00:02:37
some good in the world through data
00:02:39
science so those are our four if you'd
00:02:41
like populations of interest if you'd
00:02:43
like researchers students companies and
00:02:47
the community and Now's the Time to have
00:02:49
that impact I love that let's talk about
00:02:52
a couple of those buckets let's start
00:02:53
with number two which you mentioned
00:02:55
students now this initiative you've got
00:02:57
this really unique collaboration with
00:02:59
open AI most of us know that name by now
00:03:01
because they're the makers of chat gbt
00:03:03
so starting in the fall our MBA students
00:03:06
are going to have these creative
00:03:08
Enterprise licenses with chat PT what do
00:03:11
those licenses do what are they going to
00:03:13
create with them what's the end result
00:03:15
yeah so first let me just say why I
00:03:17
think this is crucially important and
00:03:19
let me just say as you can imagine
00:03:21
there's more sensitivities on younger
00:03:23
adults meaning the undergraduate
00:03:25
population but that's coming by the way
00:03:27
there will be a time where Wharton will
00:03:28
be able to present enter priz licenses
00:03:30
to the entire population but of course
00:03:32
the undergraduate is as part of the
00:03:34
University of Pennsylvania so there's
00:03:35
some technicalities there but we're
00:03:37
working through them the the answer is
00:03:40
Andrew you learn AI by doing it right so
00:03:45
that's the only way to get people to do
00:03:48
it every warden for now MBA every
00:03:50
faculty member every PhD student will
00:03:53
have access to if you'd like the newest
00:03:55
versions as they come of chat GPT right
00:03:58
on their desktop or pH they can actually
00:04:01
experiment with it see what it can do to
00:04:03
enhance their education we don't view it
00:04:05
as replacing us as faculty we view it as
00:04:08
enhancing their education and at the end
00:04:10
of the day this is what it's going to be
00:04:12
expected of all of our students when
00:04:14
they go into the workforce they're going
00:04:16
to be expected to know so how do I
00:04:18
create video using artificial
00:04:20
intelligence how do I do the proper
00:04:23
prompt engineering so that I can get the
00:04:25
right answers out of these large
00:04:27
language model engines like chat GPT and
00:04:30
the only way to do it is to experience
00:04:33
it and that's what these licens are
00:04:34
going to provide and the good news about
00:04:36
the Enterprise version is the data is
00:04:39
secure what I mean by that is if Eric
00:04:42
bradow myself Professor bradow uploads
00:04:45
his course materials and creates a
00:04:48
localized version of chat GPT it's not
00:04:51
being sucked up into the Corpus that
00:04:54
they're using to train the model so
00:04:56
people around the world do not have
00:04:58
access to my information and let's be
00:05:00
honest Angie that's the business model
00:05:02
because chat GPT for most people is free
00:05:05
forget the Enterprise it's free it's
00:05:07
like Google Google searches free to you
00:05:10
and me but it's not free to companies
00:05:11
that want to advertise on it chat jpt is
00:05:14
free to you and me but it's not free to
00:05:18
Amazon that wants to create a localized
00:05:20
version of it so that's the way it's
00:05:22
going to work we Wharton have invested
00:05:24
money and resources and an Enterprise
00:05:26
license so we can educate our students
00:05:30
on how to be leaders through using
00:05:32
artificial intelligence I hope one of
00:05:34
them comes up with the next big idea
00:05:35
we'll stay tuned about that I do too
00:05:37
well there's so much research that's
00:05:39
coming out of Wharton about AI uh can
00:05:41
you give us like a quick touch on some
00:05:43
of the areas that have already been
00:05:45
impacted by our research and then what
00:05:46
you find to be the most promising yeah I
00:05:49
think research comes in a few different
00:05:51
buckets so one of the buckets are there
00:05:54
are people even though we're business
00:05:55
school there are people at Wharton that
00:05:57
are statisticians and computer
00:05:58
scientists like myself who we do work on
00:06:01
the actual let's call it the
00:06:02
mathematical models that underly
00:06:05
artificial intelligence and just so all
00:06:06
of our listeners know ai's been around
00:06:09
for about 60 years so as a PhD student I
00:06:12
learned about artificial intelligence
00:06:14
now large language models which is
00:06:17
taking a big Corpus of information
00:06:20
training them in real time at scale and
00:06:22
providing generative AI That's new but
00:06:24
the problem of predicting the next word
00:06:26
or token that's going to be used that's
00:06:28
that's not a new problem but number one
00:06:30
there's academics who are in my narrow
00:06:32
slice who are what we call
00:06:34
methodologists that will work on the
00:06:36
methods the larger bucket the much
00:06:39
larger bucket for researchers are people
00:06:41
that are going to apply AI so for
00:06:43
example how do you use P patient records
00:06:46
I'm in the Healthcare Management
00:06:47
Department here at the Wharton School I
00:06:49
partner with pen medicine I see doctors
00:06:51
writing handwritten notes on patient
00:06:54
records well those used to have to be
00:06:55
human coded now they don't I'm in the
00:06:58
management department Department I want
00:07:00
to know how AI is going to affect
00:07:01
teamwork and how it's going to affect
00:07:03
like if I can just ask a chat agent and
00:07:05
not my colleague well what does that do
00:07:07
to camaraderie and building
00:07:09
relationships with your team and so
00:07:11
there's researchers many of us that also
00:07:13
work on applications of artificial
00:07:16
intelligence and then of course there's
00:07:18
the student piece which is you know how
00:07:20
do I get my job done how do I get my job
00:07:22
as a student done using artificial
00:07:25
intelligence so I view US almost like as
00:07:27
a three prong at least three prong lab
00:07:29
laboratory which is the methods part the
00:07:32
applications part and the how do I get
00:07:34
it done part with artificial
00:07:36
intelligence you know I have been
00:07:38
learning so much about AI just from
00:07:40
coverage at Wharton and there really is
00:07:43
no area that it doesn't touch it really
00:07:45
isn't and I think one of the things
00:07:47
that's really um great about the school
00:07:50
is that our professors are always trying
00:07:52
to pursue relevant research research
00:07:54
that has a business case a use case can
00:07:57
you tell me uh what the companies are
00:07:59
asking us for right now what is it that
00:08:01
they want in terms of AI research yeah
00:08:04
there's there's a lot of it so let me
00:08:05
just give you a couple of specific use
00:08:07
cases so obviously one big area is in
00:08:11
customer service so for example can AI
00:08:13
be used for chatbots and what's people's
00:08:15
reaction to the use of AI generated
00:08:18
chatbots can AI be used to coordinate or
00:08:22
collect you know let's call it audio
00:08:24
recordings and video recordings make
00:08:26
them searchable indexable so that in
00:08:29
some sense we don't need humans taking
00:08:31
notes the AI engine can do it the third
00:08:33
part is what most companies are
00:08:34
interested in is they have data all over
00:08:36
the company and I say data I don't mean
00:08:39
just numeric data which is historically
00:08:42
how we thought about it but I have
00:08:44
physical reports I have online reports
00:08:47
I've got Excel spreadsheets I've got
00:08:50
PowerPoint decks I've got a SQL database
00:08:52
I've got a big data Lake sitting
00:08:54
somewhere how do I bring all of that
00:08:56
information together so Angie sitting at
00:08:59
her desk doesn't need to be a programmer
00:09:01
to query this database can just use the
00:09:04
English language or even a foreign
00:09:06
language to just query this database and
00:09:09
say what do we know about the following
00:09:11
topic if we were to make an investment
00:09:14
in X what does our data say about this
00:09:17
or Angie has a data set you're not a you
00:09:20
don't know python but you want to do
00:09:21
some analysis well you can now do that
00:09:24
analysis so I think of it as are there
00:09:26
jobs that will be replaced absolutely
00:09:29
are there but that's not to me the large
00:09:32
part the large part is how does each
00:09:34
person use AI in their job as my
00:09:37
colleague Stephano ponton says and my
00:09:39
colleague Ethan mik say it's Ai and the
00:09:42
workforce not AI or the workforce right
00:09:45
so that's what I see as the biggest most
00:09:47
exciting use cases and that's something
00:09:50
I you know I think is really important
00:09:52
to touch on too because AI has been
00:09:54
described as everything from society
00:09:56
salvation to its Doom right it just sort
00:09:58
of depends on where you sit on the
00:09:59
Spectrum as a Communications person I
00:10:01
think I lean more towards the Doom part
00:10:04
tell me tell me where you said you know
00:10:06
what what gives you optimism about Ai
00:10:08
and what do you worry about yeah so um I
00:10:13
actually given we have a political
00:10:15
election here in the US in 2024 my
00:10:18
worries are more on a personal level
00:10:20
which is around
00:10:22
disinformation and you know if you like
00:10:25
creating fake content which is almost
00:10:27
indistinguishable if not totally IND
00:10:29
distinguishable from real content I'm
00:10:31
worried about um I was actually at a
00:10:33
conference just yesterday in San
00:10:35
Francisco or two days ago I took a red
00:10:37
eye back um and this was someone from
00:10:40
one of the large security companies in
00:10:41
the world and says imagine the following
00:10:44
imagine you create a million fake
00:10:46
articles and load them into chat GPT
00:10:49
well it's going to train on those fake
00:10:50
articles and then to anybody using chat
00:10:52
GPT it's going to seem like the truth so
00:10:55
the
00:10:56
disinformation and the fact is sometimes
00:10:58
as you know
00:11:00
they call them fakes it just makes stuff
00:11:01
up it will give you an answer so that
00:11:04
concerns me in a society that's got some
00:11:07
polarization in it and a society where
00:11:10
we're relying on these engines to
00:11:13
provide us answers and my view is
00:11:15
relying is the wrong word we shouldn't
00:11:17
be relying on them they should be
00:11:19
decision support tools but I can promise
00:11:21
you I'm not betting my company's future
00:11:24
or my academic future or my teaching
00:11:27
future or my research future on what an
00:11:29
AI engine says because I know I test it
00:11:32
all the time some of it stuff it says is
00:11:34
just wrong it seems right but it's
00:11:36
actually wrong so my concerns are around
00:11:40
disinformation I've look I everybody I
00:11:42
love action movies I've seen every
00:11:44
Schwarzenegger Terminator movie I'm not
00:11:47
worried about AI cyber whatever they
00:11:49
called it cyber whatever the name of the
00:11:51
thing was in in The Terminator movies
00:11:53
I'm not worried about AI taking over all
00:11:55
of that about replacement yeah I'm not
00:11:58
I'm not worried about that but I do
00:12:00
think and I'll tell you what else I'm
00:12:01
worried about I'm worried about people
00:12:03
I'm not worried about the young people
00:12:04
today because they always get trained in
00:12:07
the best of technology and we have to
00:12:08
democratize that we need to do better
00:12:10
job of making access available I'm
00:12:12
worried about the 50 plus year old I'm
00:12:14
pointing it myself same who need be
00:12:18
training because of what AI can do and
00:12:21
where do they go so I'm worried that the
00:12:24
people whose jobs will be displaced many
00:12:27
of them don't have the training or won't
00:12:29
receive the training to make AI an asset
00:12:32
for them it'll be seen as taking over
00:12:34
their job so those if you ask me my
00:12:36
areas of concern it's I'll call it
00:12:39
disinformation and it's also that um in
00:12:42
some sense the people that are going to
00:12:44
benefit the most from it are the people
00:12:46
that are the young people who are going
00:12:47
to get training but what about the 40 50
00:12:49
60 70 plus year olds in the workforce
00:12:52
that will need this type of training
00:12:54
where are they going to go to get it
00:12:56
right and they already face a certain
00:12:58
degree of age discrimin in general so
00:13:00
this might further set them back in the
00:13:01
workforce well let's not end on a on a
00:13:04
tough note let's end on a positive note
00:13:05
what gives you optimism about AI for the
00:13:08
future and about this
00:13:09
initiative well I I think I hope this is
00:13:12
what everybody does and my my colleague
00:13:14
Ethan mik says this the best just go on
00:13:17
and start using it like think about all
00:13:19
the things you do so I'll give you an
00:13:21
example I'll just take an example matter
00:13:23
of fact our our interim president Larry
00:13:25
Jameson asked me to speak on this to a
00:13:27
group of uh if you'd like alum I maybe
00:13:29
about two weeks ago the role of AI and
00:13:31
education and I said okay so I used to
00:13:34
need I still do I could use them but I
00:13:37
don't need them physical Tas in the
00:13:39
classroom to sit there and record class
00:13:41
participation well I don't need that I
00:13:42
have audio data well I needed people to
00:13:45
help me do grading it's not obvious I
00:13:47
need that anymore um well every year oh
00:13:50
God I gotta create an exam no I don't
00:13:53
chat GPT can create an exam for me oh I
00:13:56
need to create a simulation for the
00:13:58
class
00:13:59
well actually um the generative AI lab
00:14:02
that Ethan and his wife lck mik are
00:14:04
starting can generate a simulation for
00:14:06
me so the way I view it is Wharton
00:14:09
should be thrilled Eric bradow gets to
00:14:11
spend more time on Research he gets to
00:14:14
spend more time teaching in a
00:14:16
pedagogically interesting way he gets to
00:14:19
spend more time mentoring our students
00:14:21
but all this stuff that now can be his
00:14:24
time can be alleviated because he can
00:14:26
get the generative AI engine to do it he
00:14:29
doesn't have to spend his time on
00:14:30
anymore so to me I think this is I think
00:14:33
everybody this is my fatherly advice to
00:14:36
everybody here think about the tasks
00:14:39
that you do and just do it before you do
00:14:42
the tasks in the humanlike manner you've
00:14:44
been doing them till now try doing them
00:14:46
in a generative AI engine a large
00:14:48
language mod engine see the results you
00:14:50
get and step by step you'll notice in
00:14:53
yourself like I did replacing the things
00:14:55
that used to take me a ton of time for
00:14:58
weren't what I call the heavy brain
00:15:00
human CPU stuff I'd like to spend my
00:15:03
time on that's what gets me excited yeah
00:15:05
that's something I have heard our
00:15:07
professors uh repeat quite often which
00:15:09
is think of think of it as a tool that
00:15:11
can free you up from the banal so that
00:15:13
you can do the more exciting aspects of
00:15:15
the job that you love so it sounds
00:15:17
sounds good to me well it sounds good to
00:15:20
me too and as I said as you said at the
00:15:22
beginning um Morton Ai and analytics
00:15:25
initiative you know uh were
00:15:27
extraordinarily ambitious we want to
00:15:29
impact the world we want to impact
00:15:31
Society businesses researchers students
00:15:36
and we're going to do it and and then
00:15:38
real magic to it is through Partnerships
00:15:41
so we partner with schools to deliver
00:15:43
education we partner with nonprofits to
00:15:46
teach nonprofits how to use them
00:15:48
obviously we partner with Scholars
00:15:49
around the world we partner with
00:15:50
companies in the sea suite and so this
00:15:53
as a matter of fact this is a time where
00:15:55
the Ivory Tower better not be the Ivory
00:15:57
Tower because we desperately need the
00:16:01
application and the corporate Partners
00:16:03
to actually bring what we're doing to
00:16:05
life so again I've never been more
00:16:07
excited and I I'm honored that our Dean
00:16:09
asked me to lead this initiative
00:16:11
congratulations to you and thank you for
00:16:13
your time today I appreciate it Eric
00:16:15
bradow is the new Vice dean of Wharton's
00:16:18
Ai and analytics initiative you can
00:16:20
learn more about the initiative sign up
00:16:22
for updates insights at their website
00:16:24
which is aien Analytics won. up.edu
00:16:29
I'm G to say that one more time aien
00:16:31
analytics won. up.edu for knowledge at
00:16:35
Wharton I'm Angie Bas thanks for joining
00:16:37
us for more insight from knowledge at
00:16:40
Wharton please visit knowledge. won.
00:16:43
up.edu
00:16:46
[Music]
