00:00:00
There are so many aspects of AI and the
00:00:03
AI story still to be written. And to
00:00:05
that end, we're going to spend time uh
00:00:07
every month on our first podcast of the
00:00:10
month talking about various aspects of
00:00:13
the AI technology and its connection to
00:00:15
us. Our guest will be Stephano Pantoni,
00:00:18
marketing professor here at the Wharton
00:00:20
School and also co-director of the
00:00:21
Wharton Human AI research program.
00:00:25
Stephano, great to talk to you. How are
00:00:26
you today, sir?
00:00:27
>> I'm great. Great to be here. excited
00:00:29
about this new segment.
00:00:30
>> Thank you, sir. Yeah, it's this is
00:00:32
obviously so important uh in terms of uh
00:00:35
what we're seeing going on in our lives.
00:00:37
A couple of areas that we wanted to
00:00:38
focus on today. First, how AI and
00:00:41
platforms will be adjusting to the needs
00:00:44
of the consumer. And one of the things I
00:00:46
know you're focused on is the potential
00:00:48
of monetization of AI platforms. give us
00:00:51
a little uh explanation as to what we
00:00:54
might be looking forward to
00:00:56
>> in a segment like this one. Trying to
00:00:58
dissect uh the news and what they mean
00:01:02
uh in this field of AI is tricky because
00:01:05
there are so many there's so many things
00:01:07
happening all the time. But uh to you
00:01:09
know pick a couple of things that might
00:01:11
be interesting to our audience. one that
00:01:14
uh I think has come to the four uh
00:01:17
increasingly in the coming uh weeks and
00:01:19
months has been this conversation about
00:01:22
uh the monetization of chat bots. And if
00:01:26
you look at all large AI uh labs, they
00:01:30
have basically pursued so far what you
00:01:32
could call a premium strategy where
00:01:34
people can sign up and make an account
00:01:36
for Chach PT or you know Gemini or
00:01:40
Antropica or whatever. And you can do
00:01:42
that for free basically with minimal
00:01:44
hurdles. Just leave your email account
00:01:47
you confirm the email and you're in
00:01:48
business. And then paired with that you
00:01:51
have a premium subscriber strategy where
00:01:54
you would have say for example Chach
00:01:57
Plus which would set you off about 20
00:01:59
bucks a month or GP Chach PT Pro which
00:02:03
is uh the more expensive one which is I
00:02:06
believe somewhere around a couple
00:02:07
hundred a month and so so far the AI
00:02:10
labs have been able to attract very vast
00:02:13
audiences. If you look at the subscriber
00:02:15
base for chacht is enormous but a very
00:02:18
tiny percentage of the people using
00:02:20
chacht are paying for it. And uh while
00:02:24
for the eyabs uh that large user base is
00:02:27
crucial in improving the models as you
00:02:30
get increasing amount of feedback from
00:02:33
users you can make those models better.
00:02:36
But at the same time uh pressure will
00:02:39
mount on AI companies to uh uh create
00:02:43
revenues. Up to now has been uh uh you
00:02:46
know enormous amount of private capital
00:02:48
uh entering the space. So these
00:02:50
companies have been very well funded
00:02:52
without having to worry too much about
00:02:54
uh revenues. But that may change over
00:02:57
time as uh the industry matures and
00:03:01
maybe some of the froth um kind of like
00:03:03
starts to fade. And so basically
00:03:05
companies will have to think about these
00:03:07
free products that they're offering. How
00:03:09
am I going to we're going to make some
00:03:11
money out of that?
00:03:13
>> But will it be necessary for companies
00:03:15
to have both versions of that concept of
00:03:18
having a free model and a paid model as
00:03:20
we move forward? And and how should the
00:03:23
consumer think about the fact that
00:03:25
here's another kind of technology that
00:03:27
is being monetized and and really
00:03:30
forcing them to pay for its use? In many
00:03:32
cases
00:03:33
>> many many people believe that uh most of
00:03:36
the money in the uh AI space is going to
00:03:38
be made out of enterprise solution more
00:03:41
than consumer solution. So one pathway
00:03:44
for AI companies would be that of
00:03:46
focusing on the enterprise market. So
00:03:48
you could imagine OpenAI investing much
00:03:51
more on the uh let's say building a um
00:03:56
you know B2B organization selling
00:03:58
solutions to corporates and kind of uh
00:04:01
um play down or even wind down the
00:04:04
consumer part if that turn out to not
00:04:06
generate the kind of revenues that they
00:04:08
need. Companies like Microsoft which is
00:04:10
very much focused on the corporate
00:04:12
market will be focused there. But many
00:04:15
also make the argument and I've heard
00:04:16
that argument made by people at AI labs
00:04:19
too that the consumer market is
00:04:21
important because it creates a sight
00:04:23
guest and the interest in the technology
00:04:24
that broadly people in society are aware
00:04:26
and excited about. That's important for
00:04:28
a variety of reasons including
00:04:29
regulation, investor relations and all
00:04:31
that. But also maybe more importantly by
00:04:34
having millions and millions of people
00:04:36
interacting with the system companies
00:04:38
can uh use the data that they gather
00:04:40
through those interactions to make the
00:04:41
systems better. So a company like OpenAI
00:04:44
probably should be expected to pursue a
00:04:46
hybrid strategy where they do pursue the
00:04:48
B2B market but they also keep uh
00:04:51
pursuing the B2C market. But the
00:04:53
question is how do I make money out of
00:04:55
the B2C market? And typically what has
00:04:58
happened with digital products over the
00:05:00
last two three decades that has meant
00:05:03
advertising.
00:05:04
>> One of the other areas I know you're
00:05:06
focused on is that of energy supply and
00:05:08
that is an area uh which obviously is
00:05:11
drawing a lot of attention right now
00:05:12
because of the expected buildout that
00:05:14
we're going to have to see in order to
00:05:16
have enough energy. But there is a
00:05:19
component of that that obviously will
00:05:20
play to the consumer as well.
00:05:23
>> Yeah. So we'll uh let's uh let me final
00:05:27
make a final point about the
00:05:28
monetization part and then I I'll come
00:05:30
to this energy thing because there's
00:05:32
been a recent newspaper that is quite
00:05:34
interesting in that space. Now if uh
00:05:36
companies will be under pressure to uh
00:05:39
raise revenue through advertising then
00:05:42
that means that uh we will be seeing the
00:05:44
arrival of sponsored bots. Basically,
00:05:46
this would be bots that would be paid by
00:05:49
advertisers to promote particular
00:05:51
brands. And you can do that in a variety
00:05:53
of ways, but we'll see moves in that
00:05:55
direction. A couple of weeks ago, Elon
00:05:57
Musk, for example, announced that Grock
00:06:00
will be uh selling advertising services.
00:06:03
has been s sort of kind of weak we vague
00:06:06
into how they exactly going to do that
00:06:09
but quoting from the financial times
00:06:11
said if a user is typing to solve a
00:06:14
problem by asking rock then advertising
00:06:17
the specific solution would be ideal at
00:06:19
that point in time so you see
00:06:21
potentially a value of a chatbot as an
00:06:23
interface for advertisers to reach
00:06:25
audiences the interesting spin on that
00:06:27
would be from the point of the consumer
00:06:29
h do I trust a bot if I know that the
00:06:32
answers might be sponsored and how do we
00:06:34
navigate that trust issue as an AI lab
00:06:37
and potentially also regulators might be
00:06:39
concerned about disclosure and
00:06:41
deception. And so I think this idea of
00:06:43
advertising as a revenue engine for the
00:06:47
free product of chatbots and AI
00:06:50
companies I think brings up a lot of
00:06:52
really interesting and complicated
00:06:54
questions around consumer welfare trust
00:06:57
in AI companies and regulation. Now
00:07:00
moving to uh the uh energy question you
00:07:03
were asking. Um just about a week ago um
00:07:06
researchers at Google put out a new
00:07:08
paper where they uh reported a very uh
00:07:12
detailed in-depth analysis of the carbon
00:07:16
footprint of uh their Gemini chatbot.
00:07:20
And there is also a quite stunning in
00:07:22
that uh uh we've seen a tremendous
00:07:25
improvement in the efficiency of these
00:07:27
models. And so within a 12 months period
00:07:32
from May 24 to May 25, Google documented
00:07:36
a uh uh 44 fold decrease in car
00:07:40
footprint for for Aquery on Gemini and
00:07:43
that is enormous within a one one short
00:07:46
year. So climate concerns are real when
00:07:49
it comes to AI. We're building gigantic
00:07:51
data centers that are going to suck an
00:07:53
enormous amount of energy. But those are
00:07:55
more uh function of increasing huges of
00:07:59
AI than AI being very power hungry. In
00:08:02
fact, uh Aquarium on Gemini now already
00:08:05
today is equivalent to watching TV for 9
00:08:08
minutes sorry 9 seconds and about uh
00:08:12
five drops water consumption. So we
00:08:14
we're talking about fast shrinking
00:08:16
carbon footprints which like I said you
00:08:19
know if you grow a,000% a year even if
00:08:21
you're becoming more efficient you're
00:08:22
still using a lot more energy but I
00:08:24
think for the future we are at least
00:08:26
seeing great improvements in a way there
00:08:29
are two races in AI there's a race at
00:08:31
the top to build the biggest most
00:08:33
performing models and that's what the
00:08:35
race that takes all the attention but
00:08:37
there's another race going on a little
00:08:38
bit under the radar which is that to
00:08:39
create the most efficient cheapest model
00:08:42
that enable you to have the same outcome
00:08:43
comes with a lower carbon footprint.
00:08:46
>> Stephano, great to talk to you as
00:08:47
always. Thanks very much for your time
00:08:48
today. All the best.
00:08:50
>> Thank you, Dan.
00:08:51
>> You got it. Stephano Pantoni, who is a
00:08:53
marketing professor here at the Wharton
00:08:55
School and also co-director of the
00:08:56
Wharton Human AI research program.
