Latest / Becker’s Healthcare Podcast / Reimagining Healthcare with AI with T.Y. Alvin Liu, MD

Transcript

Reimagining Healthcare with AI with T.Y. Alvin Liu, MD · Becker’s Healthcare Podcast

0:01
This is where health care leadership comes together.
0:04
Becker's sixteenth annual meeting brings more than 3,500
0:07
hospital and health system executives and nearly 800
0:10
speakers to Chicago
0:12
April.
0:14
This year's event includes keynote conversations with Dallas
0:17
Cowboys legend Troy Aikman and former president George
0:20
w Bush. For the agenda and event details,
0:22
visit beckershospitalreview.com
0:25
and click on the events tab in the
0:26
upper right. We're looking forward to hosting you
0:29
in Chicago.
0:31
This is Scott Becker with dBecker's Healthcare Podcast.
0:35
I'm thrilled today to be joined by your
0:37
brilliant physician leader. We're joined today by doctor
0:40
Alvin Liu.
0:42
He is the endowed professor and inaugural director
0:45
of the James Gillis junior
0:47
MD and Heather Gills artificial intelligence innovation center
0:51
at Johns Hopkins Medicine.
0:53
Doctor Lu's had this remarkable career. He's a
0:56
he's a gifted retinal virtual retinal surgeon himself.
1:01
Doctor Wu, can you take a moment? You're
1:02
also an investor, PTX Capital.
1:05
Take a moment and introduce yourself, then we're
1:07
gonna talk about what you're doing with the
1:09
AI Innovation Center and a lot more. Doctor
1:12
Wu?
1:13
Hi, Scott. Thanks for having me. It's a
1:14
pleasure to be here today. As you mentioned,
1:17
I'm a retinal surgeon by training. So I
1:20
spent the majority of my adult life on
1:22
the East Coast. I attended medical school at
1:25
Columbia Columbia University in New York,
1:28
and my ophthalmology residency training
1:31
and retina surgery training, both at Johns Hopkins.
1:35
Afterwards, I stayed behind at Johns Hopkins University,
1:39
as a faculty member.
1:41
Right now, outside of my clinical practice,
1:44
my entire focus at Johns Hopkins is in
1:47
artificial intelligence.
1:49
In terms of AI, I wear two different
1:51
hats.
1:52
On the school of medicine side,
1:54
I am the director of the Gilles AI
1:56
Innovation Center,
1:58
which is the first endowed AI research center
2:00
at the school of medicine.
2:02
On the health system side, I have been
2:05
involved in the implementation
2:07
and scaling of AI tools for both clinical
2:09
and operational purposes,
2:11
such as diabetic retinopathy screening in the primary
2:14
care setting and revenue cycle management.
2:17
I also serve on the AI oversight team,
2:21
which is a leadership team with purview over
2:23
all things AI related
2:25
on the health system level.
2:27
Amazing. And and walk a bit about the
2:29
innovation center, the AI intelligence center, and and
2:32
what you're most excited about this year, where
2:34
your biggest focus is.
2:37
There
2:38
are several waves of AI that I'm watching
2:41
closely.
2:42
Maybe if you take a step back, I'd
2:43
like to provide a little bit more context.
2:46
The first two waves of AI adoption and
2:48
implementation in health care,
2:50
had been in, NV and scribing and revenue
2:53
cycle management.
2:54
And I think,
2:56
the next up and coming wave of adoption
2:59
is in voice AI.
3:01
There's currently a crop of AI native voice
3:04
AI companies
3:05
that are capable of handling inbound calls and
3:08
outbound calls and appointment scheduling.
3:11
They tend to be omnichannel
3:13
capable of both voice and voice and text.
3:17
I find these AI native voice companies very
3:19
exciting because,
3:21
we could really reimagine the entire
3:24
patient front door experience.
3:26
And these AI companies are inherently
3:28
disruptive
3:29
and could directly challenge incumbents,
3:32
whose tech stack is not AI native.
3:35
Fascinating.
3:36
And and where do you see some of
3:38
that evolving? Because, obviously, we're all watching in
3:39
the market some of the AI companies,
3:44
crush some of the software companies, at least
3:46
over the short term in the last couple
3:47
of weeks. Big software companies have taken in
3:50
the chin because the concerns that AI will
3:52
do a lot of what they do.
3:53
What's that evolution that you're watching? What what
3:55
do you see there?
3:57
That's a really interesting question. I think,
4:00
as we're as we all know, there's been
4:02
a sell off in traditional SaaS companies in
4:05
the past, couple of months. And I think
4:07
the underlying sell off is because of two
4:10
reasons.
4:11
Number one, investors are worried that
4:15
Frontier AI labs, the ones that make foundation
4:18
AI models, are moving into the application space.
4:21
And second, investors are worried that the traditional
4:25
SaaS, pricing model no no longer holds. Meaning,
4:28
in traditional SaaS model,
4:30
SaaS company will charge a platform or subscription
4:34
fee, and then they get to charge,
4:36
per human seat, for additional revenue.
4:40
I agree that these things are under threat,
4:43
but at the same time, I don't think
4:45
we will be competing,
4:48
the problem margins down to zero.
4:50
I am seeing
4:52
an emergence of AI application companies, and these
4:55
are not the hyperscalers.
4:57
These are not the foundation AI models maker.
5:01
Instead,
5:02
they are typically
5:03
companies,
5:04
led by,
5:06
leaders with deep domain expertise. They understand
5:09
very specific problems in health care and then
5:13
built amazing
5:15
AI native products on top of these,
5:17
quickly advancing language models
5:19
to solve very specific models, to solve very
5:22
specific problems with tangible ROI. So I'm not
5:26
surprised that these AI native companies
5:29
are taking the market, by storm.
5:32
At the same time, just because,
5:34
the traditional SaaS model will likely need to
5:37
change doesn't mean these,
5:39
traditional SaaS company are gonna go away. I
5:41
think they really have two options.
5:43
Either
5:44
they have to innovate very rapidly internally
5:49
to bring up you know, to really advance
5:51
their own AI capabilities,
5:53
or
5:54
they will be forced to,
5:56
buy out these AI native companies.
5:59
Regardless, I I do think that it's a
6:01
bit of a overcorrection
6:03
and,
6:05
when it comes to the the prices of
6:07
these publicly traded SaaS company.
6:10
But at the same time, we will see
6:12
some divergence in in terms of who's gonna
6:14
win it and who's gonna lose in the
6:17
next twelve to eighteen months.
6:19
Thank you so much. And and when you
6:21
look at this coming year,
6:24
what are some of the trends that you're
6:25
watching most closely,
6:28
both as a leader of an innovation AI
6:30
center and also as a physician leader?
6:33
I think there are three things that I
6:35
will be keeping track of. Number one is,
6:40
the continue
6:41
progress and traction
6:44
of health care AI companies on the app
6:46
level.
6:46
I think we'll see more and more of
6:49
these health care, AI app level companies with
6:53
very convincing products.
6:56
The market
6:57
place is going to get more and more
6:58
crowded. So at the end of the day,
7:00
I I do think that AI is getting
7:03
commoditized. The ability to build AI products
7:06
is going to become a commodity.
7:08
The name of the game, it's really implementation,
7:11
implementation, implementation.
7:13
The second trend
7:15
that I I'm quite, excited about is a
7:17
field called, oculomics.
7:20
So oculomics is the field of science, the
7:22
studies, the correlation,
7:24
between ophthalmic biomarkers, oftentimes
7:26
retinal biomarkers,
7:28
and systemic health states.
7:30
In 2026,
7:31
by combining AI and retinal images,
7:33
we could actually predict three buckets of diseases,
7:37
cardiovascular
7:38
disease, dementia, and kidney disease.
7:40
There are already commercial,
7:42
entities with scientifically validated products that could predict
7:46
these buckets of disease.
7:48
If we could deploy these retinal image and
7:52
AI based oculomic screening programs in the community,
7:56
we could significantly,
7:58
widen the patient funnel, improve,
8:00
patient access,
8:02
identify patients at risk, and really initiate,
8:05
time intervention
8:06
and improve health outcomes on a population level.
8:10
However,
8:11
this would really require buy in from different
8:13
stakeholders,
8:15
design of new patient pathways, and innovative business
8:18
models.
8:19
The third trend that I will be watching
8:21
in 2026
8:22
is the idea of
8:25
artificial
8:26
medical general intelligence.
8:28
So the definition
8:29
of artificial medical general intelligence
8:32
is still being debated, but I think
8:34
a reasonable
8:36
definition
8:37
is an AI system that is better than
8:39
75%
8:40
of physicians across all medical specialties
8:43
in terms of medical knowledge and clinical decision
8:45
making.
8:46
I think from a technology perspective,
8:49
we really are quite close to achieving this
8:51
benchmark.
8:53
How fast we can productize this ability is
8:56
a separate issue.
8:57
But I have no doubt we'll get that
8:59
there eventually.
9:00
As you can imagine,
9:02
once we achieve
9:04
artificial medical general intelligence,
9:07
medicine as we know it will be completely
9:09
transformed
9:10
and so will the business,
9:12
model of medicine.
9:14
Doctor Liu, let me ask you this question.
9:16
With the evolution in artificial intelligence,
9:19
the resources at at physician's hands
9:22
that accessible,
9:24
how will that change or should it change
9:27
medical education?
9:28
I know it's a very challenging question. Any
9:30
thoughts there on how this should impact
9:32
how we educate our physicians?
9:35
I'll answer that in two different ways. One
9:38
aspect is how we should think about AI
9:41
in the education of physicians. And the second
9:44
question is, how should we educate
9:48
current or evolving
9:50
health care executives? And I'll start with the
9:52
first one.
9:53
In my experience, in my observation,
9:56
AI in medical education, especially for physicians, is
9:59
really lagging behind. And I don't think it's
10:03
primarily
10:04
a health care medicine specific phenomenon. It's just
10:07
that AI is advancing so fast. I think
10:10
all aspects of society is playing catch up.
10:13
I will say, if I were to design
10:16
the next generation of medical education for physician,
10:20
which is start with the premise that AI
10:23
is not a fad. AI is here to
10:25
stay.
10:26
And whether we like it or not,
10:29
AI will be very tightly integrated to the
10:32
daily operations
10:33
of the provision of care and what what
10:36
it means to be a doctor. So I
10:37
think
10:38
starting with that
10:40
underlying premise
10:41
is an important thing,
10:43
for us to design the next generation of
10:46
education for physicians.
10:48
On the health care executive side,
10:51
I will say that it is clear,
10:54
that AI can improve the the efficiency of
10:57
many aspects of health care, with measurable concrete
11:01
ROIs.
11:02
That's why it's also equally important for health
11:06
care executives
11:07
to be more familiar with AI,
11:10
specifically,
11:11
which problems
11:13
could and should be solved by AI.
11:15
And, frankly, CK, a lot of problems cannot
11:18
and should not be solved by AI. So
11:20
just first and foremost,
11:22
it is important for health care executives to
11:24
understand
11:25
what AI
11:26
could and should do.
11:28
And and second,
11:29
most
11:30
current AI solutions are human centric, meaning we
11:33
try to apply AI at various points of
11:36
the workflows, which are designed by and for
11:39
humans.
11:40
Ultimately,
11:41
the efficiency gains achieved this way will be
11:44
limited and incremental.
11:46
Instead, I encourage evolving
11:48
health care exec executives,
11:51
to reimagine and redesign workflows
11:53
such that these new workflows are AI centric.
11:56
Meaning, we should design workflows to maximize the
11:59
efficiency and impact of health care systems first,
12:02
and then insert humans
12:04
wherever it makes sense.
12:06
I believe that to bring transformation changes to
12:08
health care, AI centric workflows will be critical.
12:13
Thank you. Doctor Wu, let me ask you
12:15
this question.
12:16
You've had this magnificent
12:19
professional career at the intersection of innovation,
12:23
retinal surgery,
12:25
and and just fascinating.
12:28
Talk for a second about what advice would
12:30
you give physicians
12:32
that wanna keep thriving,
12:34
that wanna have impactful careers,
12:37
and and wanna keep it exciting and interesting.
12:39
What what advice do you give to people?
12:41
I think AI is
12:44
just another technology.
12:46
If you look at
12:47
the history of medicine, it's been really in
12:50
tightly intertwined with the evolution of technology as
12:53
well. And over time, especially
12:56
over the past fifty years or so, what
12:58
it means to be a physician
13:00
and what it means to practice medicine really
13:02
has changed a lot. So I I I
13:03
don't think fundamentally that's a new thing when
13:06
it comes to AI, except that
13:08
the rate of change will only accelerate.
13:12
I would encourage
13:14
new physicians to really
13:16
think about what it means to be a
13:18
doctor. You know, broadly speaking,
13:21
it typically involves diagnosing a disease,
13:24
coming up with a treatment plan, and then
13:26
also interacting with the patient.
13:29
Certain aspects of what it means to be
13:31
a doctor will get automated or will be
13:33
heavily assisted by AI,
13:35
specifically right now in the diagnostic space. That's,
13:39
the that's an evolving landscape that's happening really
13:41
fast. I think in the near future,
13:44
AI will also play an increasingly important role
13:48
in recommending
13:49
what the optimal treatment
13:52
strategy will be. What that means is, you
13:54
know, again, what it means to be a
13:57
doctor will be different. And my prediction is
13:59
that the human aspects of being a doctor
14:02
when it comes to gaining trust, when it
14:04
comes to providing comfort when the situations get
14:06
tough will be increasingly
14:09
more important. So in a very ironic sense,
14:13
the rise of AI will force us humans
14:16
and physicians to rethink what we are really
14:19
good at And in my opinion,
14:22
fundamentally
14:23
being a physician and doctor, it's about providing
14:26
care, caring about another another human being, and
14:30
providing comfort in times of difficulty.
14:32
And I think that's what we should increasingly
14:36
focus on as well.
14:38
Thank you so much. Doctor Wu, what a
14:40
remarkable career. I wanna thank you for taking
14:43
the time to spend time with us today
14:45
on the Becker's Health Care podcast. Just remarkable
14:47
what you're doing and what Johns Hopkins is
14:49
doing. Thank you so much for joining us.
14:51
Thank Thank you, Scott, for your time. I
14:53
really enjoyed our conversation. The pleasure is mine.