00:00:00
Well, we continue our year-end series
00:00:02
today talking about artificial
00:00:04
intelligence, which obviously we have
00:00:06
seen have a growing influence on our
00:00:08
lives. But as we turn the calendar into
00:00:10
2026,
00:00:12
what should everybody think about where
00:00:14
AI is and what it is going to continue
00:00:16
to develop into into the next year?
00:00:18
Pleasure to be joined by Ethan Mollik,
00:00:20
associate professor here at the Wharton
00:00:23
School. Hi, Ethan. How are you today?
00:00:25
>> Excellent. Thanks for having me. Okay,
00:00:26
so we're coming towards the end of the
00:00:28
year and as we are doing this with this
00:00:30
variety of interviews, where do you
00:00:32
think we are right now in terms of the
00:00:34
development of AI in terms of how we use
00:00:37
it, how we think about it, and maybe
00:00:39
even more importantly, what we should
00:00:41
expect going in to the next year.
00:00:44
>> So I think the biggest news from the
00:00:46
last year is that there hasn't been a
00:00:48
slowdown in AI. I mean obviously
00:00:50
everyone's discussing about financial
00:00:52
bubbles and other concerns that we can
00:00:53
discuss but the biggest issue from
00:00:55
somebody who if you look past what the
00:00:58
ups and downs of the market are who wins
00:01:00
or who loses it's that we haven't hit
00:01:02
the wall yet that AI development
00:01:04
continues to progress that AI models
00:01:06
have gotten much smarter and better that
00:01:08
um you know for example um there about
00:01:11
2.5% of super forecasters which um
00:01:13
professor Phil Tedlock who's another
00:01:15
warden professor has been famously
00:01:16
tracking thought that the AI models
00:01:19
would get a gold in the international
00:01:20
math Olympics by 2025. That's 2.5%. Uh
00:01:23
it turned out that it um that both
00:01:25
Google and OpenAI got golds at the
00:01:27
international math Olympics this year
00:01:29
with AI models. So we're we're seeing
00:01:31
very rapid gains uh still continue and
00:01:34
from a business perspective it seems
00:01:36
like there is more and more confidence
00:01:38
by the seauite and by leadership up and
00:01:41
down the chain that AI is something that
00:01:44
they can use as an important tool moving
00:01:46
forward for their businesses on a
00:01:48
day-to-day basis. Yes. So I mean we've
00:01:50
seen surveys by fellow Wharton
00:01:52
professors on this exact topic showing
00:01:54
that a very high percentage of of seuite
00:01:58
people are suggesting that they're
00:01:59
getting returns from AI now which is a
00:02:01
big change from earlier in this year. I
00:02:02
mean I there's no sign of momentum
00:02:04
slowing down in terms of actual
00:02:06
adoption. We've hit a billion users of
00:02:08
AI on a regular basis. Again, you know,
00:02:10
the questions about business and
00:02:12
economics and environment, they're all
00:02:13
kind of separate and but intertwined
00:02:16
kinds of issues there,
00:02:17
>> right? And one of which is about labor
00:02:19
and how it's going to impact the labor
00:02:21
market. And some companies, as you know,
00:02:23
we got towards the end of 2025, uh,
00:02:26
making job cuts. A lot of companies will
00:02:28
be doing that anyway. But the question,
00:02:31
I think, for a lot of people is, will AI
00:02:33
have a significant impact on the labor
00:02:35
force moving forward? I think some
00:02:37
people say that it will work in unison
00:02:39
with employees. Other people say it
00:02:41
would replace them. I wish I had a
00:02:43
crystal ball even though this is about
00:02:45
looking forward. I mean, I think the
00:02:46
fair thing to say is that AI is probably
00:02:49
not responsible for large-scale labor
00:02:51
market changes yet as of the end of
00:02:53
2025. Although Eric Bolson and company
00:02:56
at Har at Stanford have a paper
00:02:58
suggesting that there's less juniors
00:02:59
being hired and more senior people in
00:03:01
fields like software. But it really is
00:03:03
unclear at this point. I don't think
00:03:04
there's any mass employment from AI, but
00:03:06
if you look at where the goal of the AI
00:03:08
companies are, it's to be able to build
00:03:10
as a machine smarter than a human in the
00:03:12
next couple of years to replace all
00:03:13
human labor. Uh I mean, I I think it's a
00:03:15
little unrealistic they're going to get
00:03:16
there, but we also can't count that out.
00:03:18
Um so I think we're going to see this
00:03:20
happen. I think the big debate is how do
00:03:22
we do this in a way that's proworker,
00:03:24
pro-human
00:03:25
rather than as just a way of deploying
00:03:28
or replacement for people. Looking at it
00:03:30
from a technology perspective, we
00:03:31
already are starting to hear how quantum
00:03:33
computing will have an impact on our
00:03:35
culture. But how do you see generative
00:03:37
AI continuing to to differ from a
00:03:40
variety of other technological
00:03:42
developments that may have an impact uh
00:03:44
on our society moving forward? I mean I
00:03:47
I in a lot of ways Gen AI is the sort of
00:03:49
big one that's clear. Um you know we
00:03:52
talk about quantum, we could talk about
00:03:53
fusion, but those are not those are
00:03:55
future technologies. Again, there's a
00:03:57
billion people using generative AI
00:03:58
models on a weekly basis right now,
00:04:00
right? We have evidence that they can do
00:04:02
hard level jobs, that they can do hard
00:04:04
level math, that they can, you know,
00:04:06
people are treating these things as
00:04:07
companions. They're getting help with
00:04:08
school for better, for worse. I mean,
00:04:11
that's already here. I mean, the change
00:04:12
is baked in. We could stop with AI
00:04:14
development today and have 10 years of
00:04:16
disruption as we figure stuff out. So, I
00:04:18
I always try and draw a line between
00:04:20
quantum computing, might change
00:04:21
everything, but we are still at
00:04:23
experimental quantum computers. AI is
00:04:25
deployed at scale already. Um, and you
00:04:28
know, for better or for worse, I it's
00:04:30
become a major industry.
00:04:32
>> Where then is the the growth in terms of
00:04:34
the use of of generative AI, do you
00:04:36
think in the in the next several months
00:04:38
and and years?
00:04:40
>> I mean, right now there's no sign of a
00:04:42
slowdown in the growth of AI adoption.
00:04:45
Um, it's going to move, I think,
00:04:47
increasingly inside organizations. We're
00:04:49
seeing organizational adoption increase
00:04:50
as well. and organizations are have to
00:04:52
rethink processes and approaches to
00:04:53
figure out how to use AI effectively. It
00:04:56
can't just be chat bots where people ask
00:04:58
write my essay for me or summarize this
00:05:00
data for me. It's going to have to be a
00:05:01
deeper form of interaction. So, be
00:05:03
watching for that kind of agentic work
00:05:05
starting to appear where people work
00:05:07
deeply with AI, assign AI tasks that it
00:05:09
completes and so on.
00:05:10
>> You've talked in the past about, you
00:05:12
know, certain guard rails and having
00:05:14
guardrails up around some of this. How
00:05:16
important are the guard rails uh as we
00:05:19
move forward?
00:05:21
So the guard rails um are an important
00:05:24
part of long-term AI safety and also the
00:05:27
immediate safety. We know people consult
00:05:30
AI on psychological matters, legal
00:05:31
matters, medical issues. Having the AI
00:05:34
answer accurately and correctly and tell
00:05:38
people to seek help when needed and you
00:05:40
know provide and refuse to answer when
00:05:43
it's ethical to do so remains a very
00:05:45
large problem.
00:05:46
>> Ethan, great to talk to you again.
00:05:48
Thanks very much for your time. All the
00:05:49
best. Thanks for having me. Thank you.
00:05:51
Ethan Mollik, associate professor here
00:05:53
at the Wharton School.
