00:00:00
I think the trajectory
of financial literacy has been
00:00:04
positive over time, but positive
at a somewhat glacial pace. It's
00:00:08
not improving quickly enough.
And I think recent government
00:00:12
policy and actions that we've
seen over the last four or five
00:00:15
years is a positive step
forward, but raises perhaps more
00:00:18
questions than— concerns that it
calms, right? Who's going to teach
00:00:23
it? - Right. - How's it going to be
taught? What's going to be
00:00:26
taught? There are a lot of
challenges.
00:00:29
So here we are now. If we're
moving at kind of a glacial
00:00:32
pace, here comes artificial
intelligence, which is certainly
00:00:36
not moving glacial. It's moving
about as fast as you possibly
00:00:39
can. And I guess the question
is, if AI is having all of these
00:00:43
impacts in all of these
different areas, where does it
00:00:47
potentially fit in, in and
around financial literacy?
00:00:50
Welcome to The Ripple Effect,
the podcast that takes you on a
00:00:54
journey through the minds of
Wharton faculty. I'm your host,
00:00:57
Dan Loney. And in each episode,
we'll be diving deep into the
00:01:00
inspiration behind the
groundbreaking research that
00:01:03
Wharton professors have
conducted, and exploring how
00:01:06
their findings resonate with the
world today. Well, in this world
00:01:10
where seemingly everything is
being linked to artificial
00:01:13
intelligence, we take a look now
at whether or not AI could
00:01:18
potentially have an impact on
something like financial
00:01:21
literacy. Pleasure to be joined
here in studio by Michael
00:01:24
Roberts, Finance Professor here
at the Wharton School. Great to
00:01:27
see you. - You too, Dan.
- Before we dig into the AI side of it,
00:01:30
though, I would like to get your
thoughts on where you think
00:01:33
kind of financial literacy is in
general, in our country, and
00:01:37
maybe in around the world right
now. - Sure.
00:01:39
I guess I can start off by
saying, I'm not entirely sure
00:01:41
what financial literacy even
means at this point. But if I
00:01:44
think about the research on the
subject and my engagement with
00:01:48
practitioners, I would describe
it in several ways. So first, in
00:01:52
general, it's pretty poor. If
you think about financial
00:01:55
literacy rates, they're about
57% in the US. But that hides a
00:02:01
lot of variation in the
population across genders,
00:02:04
across income, across education,
across age, where we see rates
00:02:09
varying a great deal. I think
the trajectory of financial
00:02:14
literacy has been positive over
time, but positive at a somewhat
00:02:19
glacial pace. It's not improving
quickly enough. And I think
00:02:23
recent government policy and
actions that we've seen over the
00:02:27
last four or five years is a
positive step forward. But— but
00:02:31
raises perhaps more questions
than concerns that it calms,
00:02:36
right? Who's going to teach it? - Right.
- How's it going to be
00:02:38
taught? What's going to
be taught?
00:02:40
There are a lot of challenges.
00:02:42
So here we are now. If we're
moving at kind of a— a glacial
00:02:47
pace, here comes artificial
intelligence, which is certainly
00:02:50
not moving glacial. It's moving
about as fast as you possibly
00:02:54
can. And I guess the question
is, if AI is having all of these
00:02:58
impacts in all of these
different areas, where does it
00:03:02
potentially fit in, in and
around financial literacy? - Yeah,
00:03:06
it's hard to not see it as being
central in one way or another to
00:03:11
financial literacy and financial
practice more broadly. And given
00:03:15
how quickly it's changing, as
you've mentioned, perhaps what
00:03:18
I'm about to say will be
outdated when this is eventually
00:03:20
published. Nonetheless, I'll
take a stab. Where it's at right
00:03:24
now is— is simply nowhere near
where it needs to be in terms of
00:03:29
a standalone, say, investment
advisor, or a sort of financial
00:03:34
guru for individuals. It takes a
great deal of understanding of
00:03:40
finance, and an ability to
prompt or engage with the AI to
00:03:46
get meaningful answers out of
it. I mean, when it does produce
00:03:49
answers, they can be impressive,
but they can also lead you
00:03:53
astray. So how long before we
get to a point where anyone can
00:03:57
just talk to AI and have it
solve its— their financial
00:04:00
problems? I have no idea. But
it's not gonna— it's not going to
00:04:05
negate the importance of what I
call financial proficiency.
00:04:09
You've taken a little bit of a
look at this in terms of how
00:04:13
generative AI could potentially
have a role here. What have you
00:04:16
found out?
00:04:17
I've been playing with it for
about two years, ever since— you
00:04:20
know, I had a bunch of students
in one of my class use it to
00:04:22
solve an exam. It was in a
different context, admittedly,
00:04:25
more programming and data
science. Nonetheless, it was a
00:04:28
quick wake-up call. So I've been
feeding it questions, both
00:04:33
standard financial literacy
questions, as well as questions
00:04:36
from my exams. And it's
interesting. AI is financially
00:04:41
literate by any measure, in
practice. From that perspective,
00:04:46
it's impressive. But when you
ask questions such as— you know,
00:04:51
just broad open-ended questions—
you know, "How should I save for
00:04:54
retirement?" "What should I invest
in?" Now I'm sure there's
00:04:57
regulations or restrictions on
the AI from becoming— you know, an
00:05:01
unlicensed registered RIA, but
it really— it doesn't know where
00:05:06
to begin. It provides a very
broad overview. And so what I
00:05:12
mean when I say I don't know
when it's going to be ready to
00:05:16
be our financial gurus, I don't
know how quickly AI technology
00:05:19
will progress to that point.
00:05:21
What has to happen, then,
potentially, for it to get to
00:05:24
that point— because the one
thing we do know is that as AI
00:05:28
is building up, we have to
remember there are human
00:05:30
components that are kind of
behind the scenes on all of this.
00:05:33
I think what's critical is
that I don't see AI as just a
00:05:38
panacea. I view it as a
complement and an accelerant on
00:05:42
our path towards financial
literacy, whatever that may be,
00:05:45
or financial proficiency. In
other words, I think it's going
00:05:47
to be critical that people
recognize that knowledge of
00:05:52
finance and financial principles
are not going to go away. That
00:05:57
they're going to be paramount in
order to engage successfully
00:06:02
with AI, wherever it may be in
the future.
00:06:05
But I think some people would
wonder, with the fact that we
00:06:09
know that financial literacy,
education is important, and the
00:06:13
fact that it has been a hard
thing to truly implement in the
00:06:17
right manners in schools in
different portions of the
00:06:20
country, of whether or not at
some point AI can be that
00:06:24
vehicle to be able to help teach
along the way.
00:06:28
- Oh, I think it can and it probably
will be a wonderful aid sooner
00:06:32
than we think. So Sal Khan has
created Khanmigo. - Yeah. - There's no
00:06:38
reason we can't have personal AI
TAs for our classes now, or very
00:06:46
soon. But there's a difference
between being a TA for a class,
00:06:52
and being someone who can answer
individual specific personal or
00:06:58
professional financial
questions. - Right. - That's
00:07:01
important. It's important to
recognize that distinction.
00:07:04
Everyone's different. Me saving
for retirement is different from
00:07:08
you saving for retirement, and
so on. - Right.
00:07:11
And the understanding of the
dynamics of how each person
00:07:14
feels about that, as well, are
something that maybe the AI
00:07:18
cannot fully and truly
understand. - Not
00:07:21
without the help of the end
user. And that's what I mean
00:07:23
when I say it's critical for
individuals to understand how to
00:07:27
communicate with AI. But to do
so you have to understand
00:07:30
finance. You have to know what
to ask it, and importantly, how
00:07:33
to interpret what it's spitting
back at you.
00:07:35
I noticed in the research you
had done in around AI there are
00:07:39
also instances where you kind of
rephrase some of the questions
00:07:43
that you did to try and see if
there would be a better
00:07:46
understanding for the generative
AI tool, the ChatGPT, as well.
00:07:49
Yeah, absolutely. I had to learn
how to communicate with the AI
00:07:54
in some sense in order to give
it the opportunity to get the
00:07:57
right answer. And by the same
token, I imagine the AI is
00:08:01
learning how I'm communicating
with it, in order to give me the
00:08:04
answer I'm looking for.
00:08:05
What you're saying is somewhat
of a theme that we've heard from
00:08:09
other experts, other analysts,
other business leaders— that AI
00:08:14
in the scope of where financial
literacy may be going will be a
00:08:18
path to help it along the way.
00:08:21
But it won't be the
be all end all.
00:08:23
I think it has the opportunity
to accelerate our progress. But
00:08:27
it won't be a replacement for
financial literacy,
00:08:31
financial proficiency.
00:08:33
You also talk in your work about
the ability to properly
00:08:37
interpret the answers being
given by ChatGPT as well. Yeah,
00:08:41
100%. It's not just spitting out
a number. It's giving a somewhat
00:08:47
long-winded explanation of the
answer. And so if you don't
00:08:52
understand what it's returning,
the information is not helpful.
00:08:56
It's not usable.
00:08:58
But is the expectation that,
though, there is a level of
00:09:00
development that we should
expect from ChatGPT in this
00:09:05
area and a lot of areas as we
move forward? - Yeah, I
00:09:08
think so. I mean, it's
progressing so quickly. It's
00:09:11
difficult to keep up with. And
that's why I say it's hard to
00:09:14
see a limitation, at least from
my perspective, on what it's
00:09:19
going to be able to do. Other
than I don't see it, at least
00:09:23
currently, as a replacement for
00:09:26
the progress we need to make on
the literacy side.
00:09:28
What has that research meant to
your mindset around teaching
00:09:34
finance and the component of
having ChatGPT as a tool to use
00:09:39
in that process?
00:09:40
Oh, I'm incredibly excited. I
take the view that ChatGPT or
00:09:44
other Gen AI models are just a
fantastic tool in the toolkit
00:09:50
that we need for addressing
financial challenges. So I
00:09:53
embrace it. I encourage my
students to embrace it. I want
00:09:57
them to use it on tests or
homework if they choose. But what
00:10:01
they quickly learn is that if
they're going to use it, they
00:10:04
can't shirk on understanding the
finance. - But because
00:10:07
we're talking about an area like
finance, then I think you also
00:10:11
have to bring into the
discussion the regulatory side
00:10:14
of how that industry is managed
to begin with, and potentially
00:10:18
how AI may have to be managed in
and around finance. I mean,
00:10:22
obviously I think we're talking
about a little bit larger-scale
00:10:25
discussion. But still, the
industry as a whole, there's
00:10:29
going to be a focus on how AI
kind of plays into finance as we
00:10:33
move forward.
- Absolutely.
00:10:34
It's a natural progression from
robo advisors, right? I mean,
00:10:37
just— it's a different
algorithm. That's what it is. So
00:10:40
it's going to have to be subject
to similar, though not identical
00:10:43
regulations, as we already see
in the FinTech space.
00:10:47
How much do you think, in terms
of the education piece, that
00:10:49
you've seen it— has it benefited
your students?
00:10:52
So it depends. For data science-
related financial applications,
00:10:57
when I teach my Data Science for
Finance course, it's hugely
00:11:01
helpful, because it just— it
almost— almost completely
00:11:05
eliminates the technical burden
in terms of programming and
00:11:08
working with data. It's made
things so much more efficient.
00:11:12
On the banking and personal
finance side, it's a little less
00:11:17
helpful, again, because it
struggles with broad, sort of
00:11:22
almost open-ended questions. It
varies based on application at
00:11:27
this point.
- Anything
00:11:28
that really surprised you about
kind of your interaction with
00:11:32
ChatGPT AI, as you've kind of
used it the last couple of years?
00:11:37
Yeah. Sometimes it gives
impressively insightful
00:11:42
responses. In fact, it actually—
it found an error in one of the
00:11:46
financial literacy exams, which
I thought was really impressive.
00:11:50
On the other hand, it also
impresses me in terms of some of
00:11:53
the silly mistakes it makes. So
that sort of volatility, that
00:11:59
needs to be worked out. And I'm
sure it will be as it learns
00:12:02
through more interaction and
data.
00:12:04
Michael, great to talk with you.
Thanks very much for joining us
00:12:06
here today. - Thank you. - You
got it. Michael Roberts, Finance
00:12:09
Professor here at the Wharton
School.
00:12:11
Thank you for listening
00:12:12
to The Ripple Effect. We hope
you found this episode
00:12:14
informative and engaging. Don't
forget to subscribe and leave us
00:12:18
a review so that we can continue
to bring you the best insight
00:12:21
from the Wharton School.
