WEBVTT

00:00:00.000 --> 00:00:02.585
We are here at AWS re:Invent,

00:00:02.585 --> 00:00:05.630
and one of the most interesting
conversations we're having is about

00:00:05.630 --> 00:00:10.927
the future of medicine and clinical
AI. With that in mind, we have Dr.

00:00:10.927 --> 00:00:12.429
Liz Kah, who is with me now.

00:00:12.429 --> 00:00:17.017
She focuses on innovation
at Aidoc, the leader in clinical AI.

00:00:17.267 --> 00:00:19.019
Doctor Kah, great to have you with us.

00:00:19.019 --> 00:00:19.769
Thank you.

00:00:19.769 --> 00:00:22.772
So first and foremost, what is clinical AI?

00:00:22.814 --> 00:00:23.732
It's a great question.

00:00:23.732 --> 00:00:27.527
People do define clinical AI differently,
but really when we say clinical AI,

00:00:27.569 --> 00:00:32.157
we're talking about the clinical space
with providers and patients and having AI

00:00:32.198 --> 00:00:32.657
that can help

00:00:32.657 --> 00:00:36.828
augment the decisions of providers
to enhance the outcomes for patients.

00:00:36.911 --> 00:00:40.957
How is Aidoc
using these technologies in medicine?

00:00:41.041 --> 00:00:42.792
It’s a really great question.
The evolution of

00:00:42.792 --> 00:00:46.379
AI that we've seen over the last few years
has been incredibly exciting.

00:00:46.588 --> 00:00:50.592
And at Aidoc
we take an intelligent design approach

00:00:50.592 --> 00:00:53.803
to designing the
AI solutions that we provide.

00:00:53.887 --> 00:00:55.889
We start with the problem. So
what is the problem

00:00:55.889 --> 00:00:58.558
that the health system, the provider,
the patient is experiencing?

00:00:58.558 --> 00:01:03.313
And then we try to build the best solution
to solve that problem that we can.

00:01:03.396 --> 00:01:07.817
Often that's actually a multimodal
AI solution and we're happy to announce

00:01:07.817 --> 00:01:12.072
that we just committed $30 million
to building a new foundation model

00:01:12.113 --> 00:01:15.450
which should enable us to develop
AI algorithms faster,

00:01:15.658 --> 00:01:18.036
to expand our medical use cases,

00:01:18.036 --> 00:01:21.664
and to really change the benchmark
for precision in clinical AI.

00:01:21.790 --> 00:01:26.002
And at the end of the day,
this makes doctor’s jobs easier?

00:01:26.169 --> 00:01:27.420
Yeah, right. It augments

00:01:27.420 --> 00:01:30.423
the physicians we aid the doctors
in the work that they're doing.

00:01:30.507 --> 00:01:33.343
So that's where your name came from? Yes.

00:01:33.384 --> 00:01:36.805
What challenges
is the industry facing right now?

00:01:36.805 --> 00:01:38.723
I imagine that they're numerous.

00:01:38.723 --> 00:01:41.518
Yes, there are many, many challenges.

00:01:41.518 --> 00:01:44.687
One that we see a lot and we've
heard a lot about lately is fragmentation.

00:01:44.729 --> 00:01:49.526
So there might be one company
that develops a great algorithm.

00:01:49.526 --> 00:01:52.904
That's a point solution for a problem
a hospital is experiencing.

00:01:52.904 --> 00:01:53.530
And then

00:01:53.530 --> 00:01:57.283
that company integrates with that hospital's
systems and provides the solution.

00:01:57.283 --> 00:01:58.284
And then there's another company

00:01:58.284 --> 00:02:00.745
that does the same thing, another company
that does the same thing.

00:02:00.745 --> 00:02:03.123
So as you can imagine,
there's a lot of fragmentation,

00:02:03.123 --> 00:02:07.085
duplication of efforts, redundancy
and also confusion, quite honestly.

00:02:07.168 --> 00:02:12.090
So what we decided to do at Aidoc
is build a platform for clinical AI.

00:02:12.090 --> 00:02:15.635
So we call it our AIOS,
and we invested a lot of time and money

00:02:15.802 --> 00:02:19.848
into building a platform that could host
all different types of clinical AI.

00:02:19.848 --> 00:02:24.769
So whatever AI the hospital wants to use,
we can host it with one single integration

00:02:24.769 --> 00:02:28.606
on our platform infrastructure,
and also enables us to be able to leverage

00:02:28.606 --> 00:02:30.150
third party algorithms as well.

00:02:30.150 --> 00:02:34.612
So if there's an algorithm
that exists that someone else has that’s

00:02:34.612 --> 00:02:39.325
best in class, we don't necessarily
need to create that algorithm ourselves.

00:02:39.325 --> 00:02:41.619
We can host their algorithm
on our platform.

00:02:41.619 --> 00:02:44.622
So I think the future is really about less
fragmentation,

00:02:44.622 --> 00:02:47.792
more consolidation,
and more partnership and collaboration.

00:02:47.876 --> 00:02:49.711
That sounds like a great vision.

00:02:49.711 --> 00:02:52.297
Thank you very much. Dr. Liz Kah. Thanks.