00:00:00
Well, we certainly know that there is a
00:00:02
lot of interest around AI both from the
00:00:04
public and from companies. But what's
00:00:07
interesting is about the three main
00:00:09
programs that are out there and the
00:00:12
design as to how they are supposed to be
00:00:14
best effective. Pleasure to be joined by
00:00:17
Stefano Pontoni, professor of marketing
00:00:20
here at the Wharton School who joins us
00:00:21
today from California on a foggy morning
00:00:25
out in California. Stefano, great to
00:00:27
talk to you today. How are you, sir?
00:00:29
I'm great. Thanks, Dan.
00:00:31
So, I mentioned we know that
00:00:35
the companies are wide in terms of using
00:00:38
it and varied in terms of how they're
00:00:40
using and building AI. But there is also
00:00:44
different kind of strategies that are
00:00:47
really in play here for these platforms.
00:00:50
Yeah, there's a lot of talk about about
00:00:53
not just what kind of algorithms
00:00:55
companies are building and the
00:00:57
architecture that they're using and the
00:00:59
data that they're using and all of that.
00:01:01
There's also interesting conversation to
00:01:02
be had about the business strategy that
00:01:05
different companies, different AI labs
00:01:08
are using. So, sure.
00:01:10
There's three main ones that we know
00:01:14
about ChatGPT
00:01:18
that is the main one seemingly right
00:01:20
now.
00:01:22
It seems like that to degree ChatGPT
00:01:26
is a case of first mover advantage,
00:01:28
correct?
00:01:29
Yes, so we have indeed three main
00:01:32
US-based AI labs. The first one is
00:01:35
OpenAI, like you said, the ChatGPT is a
00:01:38
product that
00:01:40
you know, basically made the generative
00:01:43
AI
00:01:45
revolution kick off in a way in a big
00:01:47
way, you can say.
00:01:49
And then you have Google which
00:01:53
is a company that invented the
00:01:54
transformer model that is powering all
00:01:57
of these generative AI tools, the
00:01:59
language-based ones. So,
00:02:01
they are very much
00:02:03
you know, at the forefront too with the
00:02:05
chatbot Gemini and other tools that
00:02:08
they're embedding in various products.
00:02:10
And also Anthropic with its
00:02:13
chatbot Claude has also been very, very
00:02:16
prominent and very successful. On top of
00:02:18
that, you have players in China, you
00:02:21
have
00:02:22
you know,
00:02:23
other other players. But I think those
00:02:26
three are indeed the main ones. Now,
00:02:28
ChatGPT indeed was the first mover and
00:02:31
that enabled them to build an enormous
00:02:33
customer base. We're talking about over
00:02:35
a billion people of active users. So,
00:02:38
the massive use user base. But the other
00:02:42
ones aren't kidding either.
00:02:46
Should we expect that as more users join
00:02:51
and try it each of these platforms that
00:02:54
the level of usage for all three will
00:02:56
somewhat level out or does the advantage
00:02:59
that ChatGPT have already
00:03:02
one that continues on for a longer
00:03:04
period of time?
00:03:05
Yeah, looking for the three main looking
00:03:08
at the strategy of the three main labs
00:03:10
in the US,
00:03:12
there's actually a working paper that
00:03:14
just came out by a team of academics
00:03:16
including Daniel McCarthy who's a
00:03:18
Wharton PhD student now professor in
00:03:20
Maryland. They're looking at mobile app
00:03:24
downloads and trying to explore the
00:03:27
niches and spaces that these different
00:03:28
companies are occupying. And what they
00:03:30
find is that so far you've had a company
00:03:33
releasing a model every few weeks, you
00:03:35
know, doing very well out of that and
00:03:37
then everybody's now excited about, you
00:03:40
know, what Google is doing, what
00:03:41
Anthropic is doing, what OpenAI is
00:03:42
doing. But that doesn't seem to have
00:03:45
come at the cost of the other companies.
00:03:47
So far, leaps in forward in terms of AI
00:03:51
capabilities have led to an expansion of
00:03:53
the pie rather than
00:03:56
then you know, direct
00:03:58
you know,
00:03:59
taking customers away from
00:04:02
from other competitors. But we'll have
00:04:03
to see as the market continues to mature
00:04:06
and you know, we reach
00:04:08
a point where almost everybody has these
00:04:11
accounts, then it becomes a little bit
00:04:12
of a matter of what kind of habits
00:04:15
people are forming, how they're getting
00:04:16
used to using this technology. And so
00:04:18
Gemini
00:04:20
is perceived to be something that will
00:04:23
be used within kind of an ecosystem,
00:04:26
correct?
00:04:27
Yes, so if you look at the three the
00:04:29
three main labs, Google is the only one
00:04:32
that you can say is
00:04:36
fully vertically integrated in so far as
00:04:38
they deal with the
00:04:40
you know,
00:04:41
the model building parts, but they also
00:04:44
are embedding AI into
00:04:48
into everything they do and also they
00:04:50
own the cloud infrastructure that you
00:04:52
need in order to power those tools.
00:04:53
Well, that is not the case or at least
00:04:55
you know, OpenAI is trying to build it,
00:04:56
but obviously it's a huge effort. And um
00:05:00
in case the case of Gemini
00:05:03
it's basically language models in
00:05:05
general will be tools that are used in
00:05:07
every Google product. They're used in
00:05:08
Gmail, they're used in Google Maps,
00:05:09
they're used everywhere. Gemini is the
00:05:11
main chatbot product and so far you've
00:05:15
seen
00:05:16
Google using it more as a
00:05:19
as a ChatGPT-like tool, more of a
00:05:22
productivity getting stuff done, also
00:05:24
images and other things. But
00:05:26
they haven't so far embedded so much
00:05:29
into their advertising which is kind of
00:05:31
interesting considering that Google's
00:05:33
business has been based on advertising.
00:05:35
In contrast what Google has been doing
00:05:37
is to build AI overviews at the top of
00:05:39
search pages. And so they are looking at
00:05:42
monetizing chatbot through
00:05:45
advertising via the search portal which
00:05:48
will change with time. We're not
00:05:49
entirely sure how. But
00:05:52
in contrast taking ChatGPT is not the
00:05:55
only product that OpenAI has basically,
00:05:57
at least the two you know, consumer
00:05:58
products. So, they are will need to
00:06:01
monetize their product directly, not as
00:06:04
part of an ecosystem the way that Google
00:06:07
is doing, but more as a standalone
00:06:08
product. And for them therefore, it's
00:06:11
very important to think about what is
00:06:12
the revenue per user that you can build
00:06:15
for that chatbot. So, they have a free
00:06:18
plan which is what you know, large
00:06:21
majority of people are using and that
00:06:22
doesn't make them any money. Like they
00:06:24
they are losing money on that free plan.
00:06:26
And then you have the paid plan with
00:06:28
different tiers and they're making money
00:06:29
on that, but there aren't enough of
00:06:31
those who convert from the free to the
00:06:34
paid plan and therefore, they announced
00:06:36
that they're going to be doing
00:06:38
advertising on ChatGPT. And so that's
00:06:41
that's an I you know, an example of how
00:06:43
the strategy pursued by the three firms
00:06:46
is getting them to make different
00:06:47
decisions. And so Claude is viewed as a
00:06:50
premium niche strategy, meaning what
00:06:53
specifically? So, Claude has a much
00:06:56
smaller number of users,
00:06:58
much smaller compared to ChatGPT but
00:07:00
also compared to Gemini.
00:07:02
However, the revenue per user is
00:07:05
extremely high and Dan McCarthy team
00:07:09
estimates that the revenue per user for
00:07:11
Claude is about 40 times higher than the
00:07:15
revenue per user from Gemini for
00:07:16
example. So, massive difference. And the
00:07:19
reason for that is that they have
00:07:21
pursued a more professional market,
00:07:24
basically coding through Claude code and
00:07:27
poor work and tools like that. And that
00:07:30
obviously is a market that is willing to
00:07:32
pay because it's so easy to justify
00:07:35
spending money on the chatbot if you
00:07:37
find it that it really helps you create
00:07:40
code with much faster speed or much
00:07:42
higher quality. So, in that sense indeed
00:07:45
it's a niche strategy. Niche not to mean
00:07:48
very small, we're still talking about
00:07:49
many millions of users and huge revenues
00:07:51
and growing fast.
00:07:53
But a little bit less focus on the
00:07:55
consumer. Although you can see that a
00:07:57
little bit changing. So, Anthropic was
00:07:59
doing for example, you know, advertising
00:08:01
during the Super Bowl. I think it
00:08:03
indicates a more of a mainstream mass
00:08:05
strategy.
00:08:07
So, do the strategies that these
00:08:09
platforms have right now and these
00:08:11
companies have for them
00:08:13
give us an idea of how each one of these
00:08:15
could potentially develop in the years
00:08:18
ahead?
00:08:19
Yeah, it's very interesting because
00:08:21
the strategies are very different and
00:08:24
yet it's not clear to me that
00:08:28
one is right and one is wrong. I think
00:08:30
probably there's going to be different
00:08:31
ways of being successful and it seems to
00:08:33
me that the strategies that the
00:08:35
companies have used is a good fit for
00:08:38
their companies. So, OpenAI is going to
00:08:40
go for this mass market and try to
00:08:43
build revenues out of the free users
00:08:46
using advertising while at the same time
00:08:48
they keep building at the high end for
00:08:50
paid subscribers and for enterprise.
00:08:53
You've got Gemini is going to be
00:08:54
basically part of a ecosystem play by
00:08:58
Google to leverage AI to build
00:09:01
a strong customer value across a set of
00:09:04
tools and platforms and monetize them
00:09:06
through advertising along the way. And
00:09:08
then you're going to have Anthropic
00:09:10
continuing this more niche strategy and
00:09:13
trying to basically nail the coding and
00:09:15
professional market. All of them are
00:09:18
being successful. All of them will
00:09:19
continue likely to be successful, but in
00:09:21
their own different way.
00:09:24
Because there's enough of a marketplace
00:09:25
for all three to kind of continue to
00:09:27
develop in their own landscape. The I
00:09:30
mean, the the market didn't exist
00:09:32
three and a half years ago. So, it's
00:09:34
still still still pretty fresh and
00:09:37
there's still room to grow in many
00:09:38
different ways. You can grow the number
00:09:41
of users, you can grow the revenue per
00:09:42
users, you can grow how much users are
00:09:46
doing with their platforms. So, we
00:09:48
expect that to continue evolving. And by
00:09:50
the way, the fact that one has a say
00:09:53
ChatGPT or Gemini subscription doesn't
00:09:55
mean that we not also want to
00:09:57
a closed subscription. This is not just
00:09:59
a for some users will be a need one
00:10:02
choose a one and go with that. But for
00:10:04
many people they'll have more. I mean
00:10:06
they're using all of them.
00:10:08
I think many will do that too. So if we
00:10:09
have if we can have five streaming
00:10:11
services we can have at least one
00:10:14
AI service, right? Right. I do that with
00:10:16
my students. I ask them you know who has
00:10:19
a Netflix subscription and everybody
00:10:21
raise a hand. I ask him who has you know
00:10:23
a Spotify or everybody has those
00:10:26
subscriptions and I think we'll end up
00:10:28
having many of these chat subscriptions
00:10:30
too. Stefano, great to talk to you.
00:10:32
Enjoy California. We'll see you back
00:10:34
here in Philadelphia.
00:10:35
Thanks, Dan. You got it. Stefano
00:10:37
Petroni, marketing professor here at the
00:10:39
Wharton School.
