June 24, 2021 • 45 mins
This week we are delighted to welcome technologist, entrepreneur, author and lecturer, Dr. Timothy Chou to the Podcast. We are going to discuss the power of the Cloud and what it could mean for children’s healthcare and research.
In his own words, Timothy has been lucky enough to have a career spanning academia, successful (and not so successful) start-ups and large corporations and he is also an author.
As President of Oracle’s original cloud business, Oracle On Demand, he grew the cloud business from it’s very beginning. Today he serves on the Board of Directors of Blackbaud and Teradata.
Timothy started his career at one of the original Kleiner Perkins start-ups, Tandem Computers. Now as the Chairman of the Alchemist Accelerator he is focused on next generation enterprise software start-ups.
Dr. Chou started teaching at Stanford University in 1982 and launched the university’s first class on cloud computing. For those of you that listened Episode 25 of Podcast, this is where Timothy met his now good friend Dr Anthony Chang and where the shoots of an idea were formed to begin a project to connect all healthcare machines in all the children’s hospitals in the world, which like the consumer Internet, may completely change children’s healthcare on the planet.
Today we discuss this story and understand more about the opportunity that Timothy and his team and trying to uncover.
Visit our shop here to purchase a copy of the Thinking of Oscar Cookbook - Made with Love or Face Coverings. THANK YOU!
Thinking of Oscar website and contact details can be found here.
Follow us on Twitter here or Instagram here.
Theme Music - ‘Mountain’
–
copyright Lisa Fitzgibbon 2000
Written & performed by Lisa Fitzgibbon,
Violin Jane Griffiths
Podcast artwork thanks to The Podcast Design Experts
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Timothy (00:00):
We like to talk about technology for good. But a lot
(00:04):
of what we do is make it easierto buy stuff and sell stuff. And
I think this mission really is amission where we can take our
brains, our resources, ourconnections, and really make a
material difference inchildren's health care.
David (00:25):
Welcome back to The not mini adults podcast pioneers for
children's health care andwellbeing. My name is David Cole
and together with my wife,Hannah, we are the co founders
of UK children's charityThinking of Oscar. This is
Episode Two of the third seasonand we are delighted to welcome
from Stanford University, Dr.
Timothy Chu. Dr. Chu, in his ownwords has been lucky enough to
(00:46):
have a career spanning academia,successful and not so successful
startups and large corporationsand also as an author. As
president of Oracle's originalcloud business Oracle on demand
he grew the business from itsvery beginning. Today he serves
on the board of directors forcompanies such as Blackboard and
Teradata. Timothy started hiscareer as one of the original
(01:08):
Kleiner Perkins startups, tandemcomputers. Now as the chairman
of the alchemists accelerator isfocused on the next generation
enterprise software startups.
Finally, he started teaching atStanford University in 1982, and
launched the university's firstclass on cloud computing. For
those of you that listened toour podcast last week, this is
(01:30):
where he met his now greatfriend, Dr. Anthony Chang. And
where the shoots have an ideawere formed to begin a project
to connect all healthcaremachines in all children's
hospitals in the world, which,like the consumer internet, may
completely change children'shealth care on the planet.
Today, we discuss this story andunderstand more about the
opportunity that Timothy and histeam are trying to uncover. We
(01:52):
really hope you enjoyed thisconversation. Timothy, Hi, thank
you so much for joining us onthe not mini adults podcast.
Timothy (02:02):
Thanks for inviting me, David.
David (02:04):
It's an absolute pleasure. I think the story that
we're about to hear and theconversation that we've that
we've gotten the topic is prettydifferent to what we've had
previously. So we're prettyexcited to share your vision,
what you're doing and know theinitiative that you're working
on. And it's all going to berevolving around clouds, right.
But I think to begin with, whenHannah and I were talking about
this conversation, the questionthat came up was, how did you
(02:27):
get into paediatrics? So can youtalk a little bit about your
kind of journey and know how yougot into doing what you're
doing?
Timothy (02:36):
Yeah. Be glad to. So my background, which is probably
unlike a lot of your guest isall coming out of IT and tech.
So I actually came to SiliconValley in the early 80s. To work
for one of the first KleinerPerkins startups, called tandem
computers, fundamentally havenever left the area, ended up
(02:57):
being the president, the firstpresident of Oracle's cloud
business. Actually, when I leftthere, I ended up starting the
first class on cloud computingat Stanford. And that is
actually the origin of how Icame to be involved or to
understand paediatric healthcare. My class and this is a
(03:18):
pre COVID conversation. So hasabout 100 people in it. About 90
of them are local and about 10are remote. So one of the remote
students reaches out and says,I'm a little old school. I like
to meet the professor. I went,Okay, cool. So we arranged to
have breakfast at Johnny's cafeon California Avenue. And I'm
sitting there and this guy walksin, I'm looking at him going, he
(03:40):
doesn't look like a regularstudent, right? sits down, come
to learn. He has an MD and MPH,an MBA, and he's chief of
paediatric cardiology atChildren's Hospital, Orange
County. And so I'm looking athim going well, why are you
talking to me? like what can Ihelp you with? So he said, Well,
actually, I think AI, Big Data,Cloud computing and medicine
(04:02):
should meet. And so I've comeback to Stanford to earn a
master's degree inbioinformatics. I love telling
this part of the story it takesAnthony three and a half years
to finish a two year programmebecause number one, he has no
idea how to code. Number two, hestill has a real job as chief of
(04:22):
paediatric cardiology. Andnumber three, he has decided
that in his 50's, to adopt an 18month old and a six month old as
a bachelor, I love tellingAnthony story because like it's
never too late for any of us.
Right. So Anthony is really myintroduction into the world of
paediatric medicine. And themore I got into it, the more I
(04:47):
was like, boy, we can help, wemeaning us in tech. We like to
talk about technology for good,but a lot of what we do is make
it easier to buy stuff and sellstuff. And I think this mission
really is a mission where we cantake our brains, our resources,
our connections, and really makea material difference in
(05:10):
Children's Healthcare.
David (05:13):
So for those that haven't listened to our previous episode
of the podcast, you're talkingabout Anthony Chang, who, who we
had on our previous episode.
He's, as you said, everythingthat you described, but also now
a data scientist, as a result ofis working with you and taking
your course. And, you know, wetalked a lot about AI and the
opportunities there and, and Ireally wanted to once we had
(05:33):
been introduced, really kind ofbring that in and kind of
continue the story and continuethat. Yeah, or just just the
story of what you know, what webegan with Anthony, and how
you've moved it forward. And Iguess there's a bit of kindred
spirits here. Because, you know,Hannah and I have been in
technology, all of our careersas well. And we strongly see the
opportunity for, you know, newtechnologies and bringing new
(05:54):
initiatives and innovation intopaediatrics. And that's exactly
what we're trying to do from acharity perspective is bringing
the future of healthcare tochildren. So continuing your
story you've touched on andwe're going to talk about
clouds, but you've started aninitiative with Antony's
support, looking at building apaediatric cloud. But before we
go into that, I think just forpotentially some of our
(06:17):
listeners, maybe you could justgive us a quick overview of what
cloud computing or what cloudis, you know, in terms of
technology, and how that couldever be relevant to healthcare.
Hannah (06:27):
And I'm so sorry, because I am going to chip in
and make it a double prongedquestion, because it was really
interesting to have the answerto David's question of cloud,
but also, why cloud? Why did youstart there, you had the meeting
of minds with yourself andAnthony, but then there was many
areas that the two of you couldhave hit on together to address?
Timothy (06:46):
Yeah, so maybe I'll kind of go down how we ended up
thinking about cloud being ananswer to the question. And
let's start with the question.
And then obviously, per David'scomment, I'll try to give a
brief tutorial on what cloudcomputing is. So what I started
to learn by virtue of beinginvolved with the community, was
just how I'll call it primitivethings are, so today, in fact,
(07:09):
this happened just last year, akid went undiagnosed with brain
cancer for over a month inSouthern California, because it
was too difficult to move an MRIimage through two hospitals and
one clinic. And I think a lot ofpeople in the world in the
paediatric world know this.
(07:32):
They're still using CD ROMs,which like, when's the last time
you saw one of those. And ofcourse, the CD ROM has a little
sticky note on it with thepassword. So guess how much
security that is? So we learnedthat I learned that story.
There's a story that Anthonytells, a little kid in Myanmar,
who he's providing consultingfor, and she's on the operating
(07:55):
room table, they want to sharethe echo with him in real time.
They can't do it. The kid dieson the operating room table.
Later on, he gets to see theecho says, Oh, I could have told
him exactly what was going on.
And then you start looking atthis globally, one of my old
friends is actually the CEO orthe former CEO of Save the
Children. And I was describingwhat we were looking at, she
(08:17):
says, Well, you know, pneumoniais the number one killer of kids
in Africa. And it's not becausethere's not low cost scanner
technology. There's nobody toread the scan. And in fact,
about a month ago, I'm spendingtime with leading paediatrician
in Rwanda. And I come todiscover there is a single one
paediatric cardiologist in theentire country, which then leads
(08:40):
obviously in the overlap withAnthony and AI, like, well
shoot, what if we could take thebest brains from Stanford,
children's Texas children,whatever, and put it in a
computer, so that you couldactually be able to diagnose
pneumonia or COVID, or whatevernumber of circumstances, when
you look at all of that, andsay, Well, what is the root
(09:04):
cause of everything from you'restill using MRI scan, you know,
CD ROMs, etc. And oh, AI isstill difficult, because and
everybody that's in the world ofAI knows this already to be
true, and particularly AI inmedicine is there's not enough
data to drive these deeplearning networks. And if you
(09:26):
don't have enough data, you'renever going to get very accurate
algorithms at all. So you lookacross this. You go well, what's
the big issue? The big issue isthat there's not data. I mean,
there's 1000s of CT scanners outthere, right. Hundreds of
ultrasounds, Gene sequencers butnothing is connected,
(09:47):
fundamentally. And so, I like tosay in our world, we cracked
through the connection problemback in 1994. We connected a
million computing machines, andthat was the beginning of the
consumer internet. And that'sthe beginning of eBay and
Netscape. And, and we all knownow at this stage, how much
(10:09):
that's completely transformedour consumer lives. And also, by
this point in time, I hadstarted to understand, there are
about 2000 healthcare machinesin every Children's Hospital in
the world. And there are about500 children's hospitals, which
turns out, that's about amillion healthcare machines. So
(10:30):
a year ago, we launched thisproject to connect all million
healthcare machines and all thechildren's hospitals in the
world, and create a digitalinfrastructure that could
transform children's healthcare. That's the problem, the
challenge we saw, kind of ourour thinking about it obviously,
(10:51):
is on a planet wide basis, whichleads to Okay, sounds like a
great vision, how the hell areyou going to do this? So if you
look at it go, Well, if I'm agene sequencer, a blood
analyzer, whatever, and I wantto get to that data, I need a
computer, we're going to call itan edge server to talk to that
(11:13):
machine. I'm also going to needa computer to talk out, for
instance, real time stream andecho as an example. So that
computer we call an edge server.
And so what we are building orwhat we have built is an edge
cloud. Okay, why do I call it anedge cloud? So I'll get to
(11:37):
David's question. So if you tookmy class, the very first day, I
tried to explain cloudcomputing, particularly what I
call compute and storage, cloudcomputing. And if you speak,
Amazon speak, I'm talking aboutwhat they call a EC2 and S3. So
what does Amazon do? Right?
Well, Amazon says, I'm going togo buy a million servers, a
(11:59):
million computers, I'm going tomanage the performance,
availability and security ofthose computers, I'm going to
deliver him in an op x model,you don't have to purchase up
front, you can pay me everymonth. And I'm going to put
those computers in 10 datacentres in the world. And what
I've just also explained to youas well, that's Microsoft Azure,
or whether that's Google Cloud,etc, that abstraction applies
(12:22):
across the board. Okay, so thatis cloud computing, I am
delivering compute and storageas a service by managing those
servers, and changing thebusiness model to saying yes,
you can buy it, frankly, by thehour by the day by the month,
what are we doing? So our edgecloud is in many ways the same
(12:47):
thing. So we're going well,we're going to buy a million
computers, we're going to managethe performance, availability
and security of those computers,we're going to deliver them an
op x model, the only thingdifferent is we're going to put
them in a million data centres,meaning right next to connected
to the ultrasound, the genesequencer, the blood analyzer,
(13:07):
etc. And the second thing we'regoing to do differently is we're
going to fill them with data. Sowe're going to fill them both
with what we call machine data.
So that could be things like,what's the serial number of the
machine? Where's it located?
What's the laser power level onthe gene sequencer, and the
other kind of data as we callit, gnomic data, so that's the
(13:27):
data the machine would generate.
So a blood analysis a genomesequence, an echocardiogram. And
based on that allowapplications, what we call edge
applications, to be able to doany number of different things
with that data, withpermissions, obviously. So it
(13:49):
could be things which are, I'mgoing to take that data, and I'm
going to stream it to a securecell service that looks like an
Instagram, we're actuallyworking with a company called
Figure One is that to aggregateit in a centre cloud to do
centre cloud, what we're gonnacall centre cloud learning, or
(14:11):
is that and we're just in themiddle of this is it to use it
to develop federated learningapplications, all of these
become possible as to what edgeapplications do. And therefore
we're only limited by people'simagination of what becomes
possible in all this.
David (14:32):
So I'm going to try and slightly simplify what I think
you've just said just forprobably from my own benefit,
but certainly for maybe some ofour listeners, but essentially
your aim is to take paediatricdata which if you took it from
one source or one hospitalwouldn't really be enough to
actually do anything useful withpotentially. Amalgamate that
(14:54):
together so that anyone anywherecan take advantage of that data.
In order to try and solve someof the, you know, the world's
biggest, biggest problems inchild health by putting it up
onto a central system, whichwhich people can gain access to.
Timothy (15:11):
we will enable just to be clear, we will enable that
what you just said, which isaggregation in a centre cloud,
we will also enable that itnever has to be aggregated. This
is an interesting thing. So Isaid the number 2000. So, if you
look, and we just ran anexperiment at Lucile Packard, so
(15:31):
if you imagine Lucile Packard abuilding, which now has 2000
servers in it, right, that allhave an interesting amount of
resource memory, GPU, CPU, allthat good stuff. And now
connected with a highperformance private 5g network,
you could imagine that what Ijust said as I just move the
(15:54):
cloud down into Lucile Packard'sbuilding, and now all the
computing that you want to do,from even from a learning
perspective, could actually alloccur locally. I mean, this
really stretches I think, wherewe are in technology. But that
is entirely possible, just tomake a point of it. It doesn't
have to be just about learning.
This could be as simple as theexample I gave you, which is,
(16:17):
there's a kid in Myanmar, I wantto share the echo image with
Anthony in Orange County, that'sall I want to do. This
infrastructure makes that easyto do. So it can be from the
very, let's call it far reachingalmost science fiction, down to
something very simple. I justwant to share a movie, like I
(16:37):
share with my friends onInstagram. Does that makes
sense?
David (16:43):
Yeah, absolutely. I think the vision is incredible. I
guess the next question is, youknow, people that know this kind
of computing, will some of thequestions will be security,
that's a fundamental thing thatpeople will be asking about. So
let's maybe go there first. Sode identified data security or
this element? How are you goingto kind of provide that and how
(17:05):
you're going to manage that?
Timothy (17:06):
Yeah. So maybe, obviously, when we launched
this, we knew security andprivacy was had to be designed
in from the beginning, weweren't going to kind of try to
add it at the end. So from anengineering perspective, we
started there, we've actuallyengineered 28 different security
features, and compare them towhat is being done at AWS, and
(17:27):
actually, we're doing more. Partof that has to do with the fact
that we've got problems that aremore difficult to solve. So in
their world, it's kind ofdifficult to walk in the
computer room and walk away witha server in our world, you could
actually walk away with aserver. So we have to engineer a
bunch more protectionmechanisms. The other part is
privacy. Kind of fortunate. Oneof my former students is a
(17:51):
privacy lawyer with 12 years ofexperience in privacy law,
particularly the GDPR. Soeverything we've done from a
privacy point of view has beenarchitected from day one to be
around GDPR, because we knew wewanted to go global. And we see
it as a superset of what HIPAcompliance is in the US. So
(18:13):
security and privacy have been aday one objective. We think we
built a world class system here.
And as we've been going out andtalking to the various hospitals
as we deploy, we're invitingthem to go, Hey, if we missed
something, we want to know aboutit. Because this is not
something we're trying to shyaway from. We think it's an
important bedrock issue. And wewant to bring the best of
(18:36):
technology to it.
Hannah (18:39):
While I'm having that side of the capability nailed
then gives you the opportunityto go global, as you just
referred to, as you were talkingin the conversation so far,
there's a topic that's come uptime and time again, in
conversations we've had on thenot mini adults podcast, and
that's been one of equality. Soyou talked about, first of all
(19:00):
you talked about, in my words,it would be accelerating, or
reducing the time it takes todiagnose an unwell child. And
that's been at the core of ourambition for our small charity.
That was one of our firststepping stones that we were
interested in, because we couldsee the promise that technology
had. But then what I've learnedsubsequently has been more
(19:23):
around what big problem wouldyou love to resolve in
healthcare time and time again,we've heard just that there are
many complex, interrelatedfactors that lead to a very
unequal provision of health. Andso an example of a first world
country versus a third worldcountry is and is an obvious
one. But to what extent were youdriven by this sort of global
(19:47):
agenda of getting a higher levelof health care to children,
regardless of where they wereliving? Was it more about speed
or diagnosis? Do you care moreabout the global agenda? What
how did those factorsinterrelate? And what's your
thoughts on that? equality ormore quickly inequality?
Timothy (20:03):
Yeah, and you know, we're all are seeing inequality
in spades right with COVID. So,no, I, I think it's because back
to what can we as technologistsdo, that can materially impact
the world has been the agenda?
And I told you when am I, one ofmy old friends is CEO of save
the children talked aboutpneumonia? You don't really have
(20:24):
to. I mean, by the way, in myopinion, you don't have to go as
far away as Africa to seeinequality. I mean, rural
California versus MetroCalifornia level of health care
is not the same. And why is itnot the same? Well, it's because
it's a manual system, based onsilos of humans and data. So if
I've got the best and brightestin Palo Alto, well, I'm lucky,
(20:48):
right? I live here. You know,you live in you know, in
Northern California, the inUkiah, whatever, you're not
going to see that you don't haveaccess to that. And I think we
know that technology can evenaccess to information, we see it
all the time. That is theconsumer internet. I mean,
that's the other thing that Ithink a lot of times, because of
(21:11):
where healthcare has come from,people have thought about
medical devices, and pharma anddrugs and all that sort of
stuff, which all have enormousmanufacturing costs and cost to
deliver. Because we all live inthe world of software, I'm going
software, the cost ofmanufacture is zero, it's your
brain power, cost to distributeis zero. So if we can enable a
(21:33):
platform and infrastructure,which allows the best and
brightest to use their brains tocreate diagnostics,
recommendation, any number ofdifferent things, we know the
cost to build is zero, we needto know the cost to deliver is
zero, we could now changechildren's health care. I don't
(21:54):
care whether you want to talkabout rural California, rural
England, or Africa, we have thecapability. It's all sitting
here in front of us. So whydon't we do it?
Hannah (22:06):
Given that the usual and oftentimes an obstacle might be
cost? And that's not the issuehere. What obstacles do you see
in front of you? And what otherhelp might you need?
Timothy (22:17):
Well, you know, I think the obstacles are not
technological. The obstacles areboth the obstacles of change of
anything, right. And as all ofus who worked in, in IT know,
right, transformation, digitaltransformation is not a trivial
thing. I don't care whatindustry you're in healthcare
(22:39):
doesn't make it any easier.
Let's say it that way. Right. Sothe obstacles of change, which
have social things in it, youknow, organisational, etc, are
all you know, and obviously,there are challenges around
security and privacy. But I haveto tell you, I feel like all of
that we have the technology toaddress, I think what we need is
(23:00):
what we all need. When you dodigital transformation, you need
the champions, right. You needthose early adopters to go
"Yeah, I see your vision. AndI'm gonna rally my troops over
to here". Anthony is the PiedPiper have this, by the way,
just to say it. I mean, he's outthere way in front of most of
(23:21):
them, talking about artificialintelligence and what it could
do in medicine. But he's anoutlier. I mean he's a real
outlier. And I think that isreally our biggest obstacle. And
just to be nice to the medicalclinical community, I tell my
world, you know, our world, Igo, you know, if I'm a
(23:41):
clinician, I live in the hereand now I have to live in the
here and now. And now this isall I have, I can't live a year
from now or five years from now,it makes no sense. And yet, what
we all live in and what we'reused to, and what we talk about,
is we talk about where thingscould be. And I think that
disconnect between how peoplethink, right is is there now
(24:05):
we're not you know, we arefinding and, you know, between
Anthony and many other people ata lot of these children's
hospitals, there are the earlyadopters, the Vanguard's the
people who are going, Yeah, wereally need to start doing this
differently. And I think ourchallenge is how do we rally
those troops, bring themtogether? galvanise this
(24:27):
mission? Whenever we meet withpeople, I always invite them to
be part of our mission. I don'tsee it as an us them question.
Yes, we're developingtechnology, etc. But it's an all
of us thing. And I think that ifwe get the right people,
galvanise them, right, motivate,help them right, we'll get here,
(24:49):
the technology part, I'll justkeep repeating. This is not
gonna be that hard. It's thenext step which will be hard.
Hannah (25:00):
Well, and the disconnect piece is interesting. Because I
imagine you're thinking aboutyour how you make it easier to
get from where we are today tothe future. So you're showing
people the way perhaps.
Timothy (25:14):
Both showing and helping, I think the beauty of
the software business is wedon't go back to it doesn't cost
anything. So every step of theway, if we can build it, I was
just talking to a systemintegrator yesterday, who wants
to be a part of this. And Isaid, I have to tell, you know,
when I go talk to people infinancial services, I can kind
(25:37):
of say, Well, let me you want toimagine the future, this is what
it could look like, I guess theyusually can do that, because
they've been living in the worldof software for a long time now,
because there is no physicalmeaning to the world they live
in. But in medicine, that'stotally not the case. And I
said, I think it's impinnchentto begin to show them this, so
(25:59):
that they can begin to graspwhat it is, and not leave it on
a PowerPoint slide or, a speechor whatever. And I think by
doing that, and you know, again,the world of software, we can
iterate at the speed of thought,I mean, you want to build it new
again, tomorrow, no problem, wecan do that, right? And get that
(26:19):
and bring people along and showthem the power of doing this. I
do believe they all realise thatthings got to change. It's just
how does that change happen?
Where does it happen? who leadsthat change? I think those are
going to be the centralquestions. And I think that's
the work, you guys are doinggreat. I mean, we need more
(26:40):
people to want to change thisthing. And then to see how
change can happen. And I thinkin that I've already seen this,
I do have the privilege ofteaching at Stanford. So I'm on
all the faculty distributionlists. So this, this quarter,
there are three AI classes beingtaught. And so I thought, just
(27:00):
for the fun of it, I'm justgoing to post into one of them.
Because they were all goingwell, we need projects we need
projects to work on. So I justposted in one of them, basically
not much information. I justsaid, How about a project to do
AI in paediatric medicine? Youknow, within 36 hours, I had
three teams of three peoplereach out, what can we do? What
(27:23):
problems should we work on? Howcan we help? You know, I think
the will is there in thecommunity in the larger
community. And I think bybuilding a platform, a way to
begin to do something aboutthis, then we can actualize that
we can take, you know, smartkids who want to work on ,who
are maybe brilliant in federatedlearning, and marrying them up
(27:46):
with clinicians who have thevision to say, this is what we
could do with it. And they Imean, I'll just give her real
credit that we've been workingwith a doctor at Children's of
Atlanta, Rito , I talked to herand I said, Well, hey Rito what
would if we had all the data inthe world? And she said,
(28:08):
paediatric cardiology paediatricecho expert and I said if we had
all the data in the world whatwould you solve? So she goes
1,2,3,4,5 .Okay, so. So now I'mlooking at 1,2,3,4,5 going, you
know, some of these words like aand and the ,I understand. But
most of these words like, whatdoes this mean? So I invited her
(28:31):
we have a team meeting everyMonday and Thursday at two
o'clock. So I said, Hey, Ritowill you come and do a, I called
it a micro MD degree? I said,you know, words, like, what is
cardiomyopathy mean? And shecame, I was so impressed, she
(28:52):
built a presentation, just forus, it was like, let me explain
this to you. Let me explain whyyou could understand it, what
the challenges are, and tons ofvideos so that we could go Oh, I
see what she means by that. Andso I think that's the other half
of this problem, which is, youknow, we need to bring the, you
(29:15):
know, ML federated learning, youknow, AI people, and we all know
how to talk tech head, but wealso need to find a way to learn
to talk to them. And some ofthat, I think, is they need to
educate us a little bit onwhat's the vocabulary that we
can start to use. And I think inthat merger of talent of the
(29:35):
domain, expertise of a Rito andthey, you know, federated
learning expertise is coming outof Stanford or, I mean, a lot of
this work actually is occurringat Oxford, by the way, right.
And we bring these communitiestogether on top of a platform
that lets them build it, youknow, we're not that far away.
David (29:56):
We're doing a lot of you know, given that we're hopefully
people are listening Listen tothis, but we're doing a lot of
nodding along with everythingthat you're saying. And I think,
you know, that's that's probablynot evident if you're listening.
So in total agreement, but oneof the things that really stood
out for me, what you just saidis, if you actually raise the
problem, and you start to, youknow, just put it out there,
(30:18):
what can you do to help us inpaediatrics, people want to do
things they want to help? Youknow, it's a it's an evocative
subject. But I think then, aswith anything you've got, you
know, as soon as it kind of goesout of the mind's eye, then you
get back to normal and peoplejust kind of carry on. So it's a
case of continuing to have it inin the mind's eye. I guess the
next question is, is the how,how are you going to, you know,
(30:41):
what do you need to do in orderto move this forward? And you've
already touched on a little bit,but you've, you've got your
first implementation you'reworking at, funnily enough,
Stanford children's, but can youjust talk us to a little bit
about what you've done already?
And then you know, what you'rehoping to do in the future? And,
and you've got a, you've got akind of vision of a domino
effect, I guess,
Timothy (31:00):
you know, our team is all people that come out of
enterprise software. So we kindof think in terms of well, how
do you get to the customer? Andhow do you grow the business, so
we're not academics. So we, weestablished a beachhead, fairly,
obviously, at Stanford, which,by the way, it wasn't even with
all the connection networks, andeverything, still was not the
(31:22):
simplest thing in the world. Sowe understand the challenges of
their organisational structures,and kind of the established ways
of thinking about things, evenat a leading institution like
Stanford, right, so. So that'sgreat, we learned a lot, we went
in, just to tell the story, wewent and installed the edge
server, it took two and a halfhours to do the installation,
(31:46):
which acts on the one hand, andit's not so bad, because we
actually have never seen anultrasound, we had no way to
test until we showed up in thein the building. But the flip
side is two and a half hours,you're going to deploy a million
of these that we just can't dothat. So we've already done an
iteration of the software andactually rereleased it onto the
(32:07):
edge server at Stanford, whichwe believe could take this
provisioning time down to like20 minutes. So we're improving
all along the way.
David (32:16):
So just quickly, what's an edge server for those that
might not know what is it youknow?
Timothy (32:22):
So the the edge server, the edge server is a computer
about the size of your hand,we're actually looking at some
that are two inches by twoinches, but the ones we're
working with right now, by thesize, your hand, which connect
to in the case of Stanfordconnect to the Philips epic
seven ultrasound, so that everyultrasound that is done that is
(32:43):
shipped to the PAC System isshipped to the edge server. And
then the edge server has twoapplications running on it. One
which looks at the gnomic data,the question you asked earlier
David, identifies that at theedge, and then actually, for
demonstration purposes, puts itinto a box account, because
there's a nice daikon viewersitting there. The other
(33:06):
application is with a thirdparty ISV named Unitsa, which
has been managing smart devicesand for people like apple and
carvanha, and whatnot. So I'veknown the founders for a long
time, we said, well, we couldbring all that technology to
managing healthcare machines,the maintenance of health care
(33:27):
machines, utilisation of healthcare machines, etc. So that's
the second application runs onit, which communicates with them
that's the which actually is anapplication that runs at AWS. So
that edge server is that device,which those applications run on,
which communicates with theexternal world. And with the
Philips, in this case, thePhilips ultrasound. So we
(33:49):
completed that it passed all thetests. Actually, right now,
specifically, we're stillworking. There's an integration,
which we built to get the syslogdata off the Philips machine,
which is all about machine data,and that we're going through
testing right now. So the nextbig step is we said, You know, I
think we know how to scaleinside the building. How do we
(34:11):
scale across institutions. Andso the next big thing as we've
called this, the vanguardproject, this is eight
hospitals, seven more. One ofthem is obviously Stanford,
seven hospitals, where we inessence, want to replicate what
happened at Stanford, to ensurethat we have built a scalable,
(34:32):
secure environment. So we are inconversations with about 15
different hospitals out there. Alot of names that you would
know, just to say it justyesterday, but being gay Sue,
which is one of the largestchildren's hospitals in Europe,
has agreed to be one of thevanguard hospitals. So our plan
(34:53):
is to identify the eight we'recovering. One of the reasons its
eight is we're covering all thecosts of doing this. So we plan
to identify the eight by the endof May, and start installations
in June. So that's our next bigstep is moving out to eight
different sites. Once we've donethat, I'm pretty sure not that
(35:15):
we know everything. But we knowa lot about what things we need
to work on for scalability,security and performance. And
we'll do another iteration testthat but I think, you know, in
this year, we will know what ittakes to deploy a million
servers.
David (35:32):
I mean, as we've kind of already said, the vision is
incredible. What you kind ofneed to do to realise that
vision, given that it's amillion servers in every
continent on earth, and the restof it seems, seems to be, you
know, quite quite a challengeand what have you. So what, what
do you envisage the futurelooking like? And when?
Timothy (35:51):
So, you know, some people ask me a question, why
children's medicine, right? Imean, doesn't adult medicine
have the same problem? Like, Oh,well, yeah, adult medicine does
have the same problem. But theidea that we would have any
prayer of doing it in everyadult hospital, I have no idea
how you'd pull that off. Now,the nice side about children's
(36:13):
medicine, I made the pointearlier, there are about 500
children's hospitals in theworld. Well, you can guess what
that means is that about 25 ofthem, that are the leaders and
you guys have been in thisspace, I think we all could
write them down. We know whothey are. So basically, the
first objective is, let's getthe 25, of which, by the way,
(36:36):
eight, are going to be in thiscategory already. Because if I
get the 25, the other 475, well,will fall in that's not going to
be that difficult. And I that'swhy this market is somewhat
unique, I think and the prayerthat we could pull this off,
because it's this type ofmarket, I think is very high. So
(36:59):
the step after the eight, Ithink will really involve we're
already having a ton ofconversations about this. There
are two angles to the next stepone is research. So what
research projects, which youcould have never done before,
are now possible, given thisinfrastructure. And so we're
(37:20):
having conversation. In fact, Ijust had one this morning, a
children's national who's beendoing a lot of work in federated
learning, is it going to befederated learning around CT
scan data for COVID prediction?
Is that what the thing is right,or what is it going to be? And
this obviously lines up with thequestion of what becomes
fundable? The one thing I say topeople is, I'm not sure how AI
(37:43):
and adult medicine gets funded.
But I'm pretty sure there's noeasy answer to AI and paediatric
medicine. No venture capitalistis going to invest in that
market. It's just not bigenough. So it's gonna have to be
a combination of, you know,foundations, government etc
(38:05):
money that's going to have tocome in just to pick on him for
a second, we did an analysis ofNIH funding. So NIH spends about
$1.8 billion a year inpaediatric medicine $1.8
billion. We tried to identifylike, how much of that is spent
(38:25):
on AI? 15 million, maybe? Andwe're stretching, right? So
there's an education cycle.
There's a lot to bring theresearch community together. And
they have responsibility in thisas well as like, well, what are
the big clinical issues? What'sthe proposal, we're working with
a team at Stanford, you couldguess in a project called
(38:48):
chamber to do aggregatedpaediatric echo data, and
looking at the potential of thatbeing funded by a number of
different sources. So I thinkthere's one whole angle, which
is research, predominantlyaround AI, which by the way, and
I tell their community this. Youknow, we have the chance in
paediatric medicine to leapfrogadult medicine, because they
(39:10):
have all the same issues. Butyou know, how are they going to
get aggregate data? How are theygoing to do federated learning?
I mean, you know, if thiscommunity gets together, we have
the potential to leapfrog whatthey're doing. So I think that's
one whole angle that we aregoing to press forward with the
other whole angle is productionapplications. So you guys come
out of my world. So we're justtalking about ISVs independent
(39:34):
software vendors, who because ofthe technology we have built,
will enable them to do thingsthey have never done before. So
just a simple example, there's acompany out there called
Nautilus amusingly a companybuilt by the guys who built the
CD ROM exchange software. Andthey've now moved their
application into the cloud. Andso we're saying hey, let's build
(39:56):
a Nautilus edge application,which if childrens of Atlanta
decides to give it permission toaccess the echo or the CT scan,
you can now do image exchangeand annotation using the centre
cloud application at Nautilus.
(40:17):
And then I mentioned as clinicalapplication to help the clinical
engineer who today is managing.
I mean, at Stanford, there's20,000 assets in the building.
And they're managing in thedisconnected state, they don't
know where the machine is, orwhether it's being used. So you
know, if you have too manymachines, do you have too few
(40:37):
machines, no one knows. Sothey're going to be more, these
are just early kind of peoplethat we've been working with
ISVs, that we want them to usethe infrastructure to develop
new applications or extend theircurrent application. So those
are the two big directions.
David (40:57):
Typically, we could literally talk about this, given
that this is kind of our area ofinterest. Very conscious, you
know, of your time. I'm verygrateful for the time that
you've given us. One of thethings that stand out is, and
it's great that you know, thiskind of thing is happening in
pedes. First, and I think we canall agree that if it if it
(41:19):
works, which we all have fullfaith, it will, then the
opportunity to scale up toadults is there and hopefully it
has been, it makes it a loteasier from that point of view.
And that's kind of exactly theessence of what we're trying to
convey for, from not mini adultspodcast perspective, I've got a
couple of kind of closingquestions for you. One is in one
(41:41):
sentence, you've talked about ita lot, but I'm just conscious
that some people won'tunderstand necessarily tell us
what federated learning is inkind of one sentence or, or
maybe two.
Timothy (41:52):
I don't know that I'm good enough to do it in one
sentence, but I'll explain itthis way. In contrast to
federated learning, classiclearning is, I take data from
many different sources. And Ibring it into a central
location. And I try to train myalgorithm using all the data in
a central place. What federatedlearning says is you don't need
(42:16):
to do that you can leave it inplace, and only transmit model
parameters. So it's all beenarchitected. In fact, if there's
a great video that Andy Traskdoes, you watch this thing, it's
all been done for healthcare,because of privacy concerns. So
rather than transmit the dataanywhere, just transmit model
(42:39):
parameters, so do the learningat the edge is how we would say
it right. So that you never haveto let the data leave the third
floor of alder Hey, just to pickon him. Right?
David (42:52):
Yeah, no, perfect. Thank you. And if anyone so if there
is someone from a hospital thatmay not be in that eight that is
listening and is interested,what do they need to do?
Timothy (43:02):
Oh, reach out. We'd love to talk. Do you guys, you
put a email address down. Justshare the details. I just as a
comment, I'm on LinkedIn toreach out. And LinkedIn is
another easy answer to this. Weare a consortium of the willing.
So anybody willing, come ondown?
David (43:23):
Well, so we will certainly share all your details
on that on the notes, soeveryone can can hopefully get
in contact. Final question thatwe ask everybody that comes on
to the podcast, Timothy, whichis if you could solve one
problem within paediatrics, whatwould it be?
Timothy (43:39):
Well, it sounds self serving. But I think we are
solving the one problem,paediatrics. You know if we
could pull this off, I don'tactually think it's a if it's
when we pull this off. It willchange everything. I mean, the
evidence I have of this is justimagine our world before 1994.
And imagine the world today,it's so completely different,
(44:01):
because we live in the connectedstate. And they live today in
that fractured siloed world thatwe all used to live in back in
the 80s. And that level ofchange is possible with
technology. And the willing andI you know, like I said,
probably self serving, but Ithink this is the one thing.
David (44:23):
Thank you. I had to ask the question. I've failed to ask
it previously, or at least tryto guess what the answer is and
got it entirely wrong but Ithought you'd say that. I'm glad
you said that. Thank you so muchfor joining us. Thank you for
your time, I think, you know,it's been a unique and
fascinating conversation, and Ihope everyone will agree with
that. So really appreciate it.
Timothy (44:44):
I appreciate you guys taking the effort to do
something like this, you know,without getting communicating
without sharing ideas is we'renot going to make it very far.
So I'll equally say thank youguys for doing these podcasts.
And I love the name Not miniadult. It's great.
David (45:06):
Thank you so much to Timothy for joining us on the
podcast and we wish him everysuccess in his endeavour, we can
really see how connecting theworld's paediatric data can have
such a huge impact on the healthof our children and children for
generations to come. Next week,we're bringing the conversation
back to the UK, and we're joinedby Dr. Don Sharkey to discuss
(45:29):
innovations and technologieswithin neonatal care. We really
hope that you can join us thenplease do subscribe to the
podcast. And if you're enjoyingit, please do leave us a review
as well. We hope you'll join usagain next week.
We like to talk about technology for good. But a lot
(00:04):
of what we do is make it easierto buy stuff and sell stuff. And
I think this mission really is amission where we can take our
brains, our resources, ourconnections, and really make a
material difference inchildren's health care.
David (00:25):
Welcome back to The not mini adults podcast pioneers for
children's health care andwellbeing. My name is David Cole
and together with my wife,Hannah, we are the co founders
of UK children's charityThinking of Oscar. This is
Episode Two of the third seasonand we are delighted to welcome
from Stanford University, Dr.
Timothy Chu. Dr. Chu, in his ownwords has been lucky enough to
(00:46):
have a career spanning academia,successful and not so successful
startups and large corporationsand also as an author. As
president of Oracle's originalcloud business Oracle on demand
he grew the business from itsvery beginning. Today he serves
on the board of directors forcompanies such as Blackboard and
Teradata. Timothy started hiscareer as one of the original
(01:08):
Kleiner Perkins startups, tandemcomputers. Now as the chairman
of the alchemists accelerator isfocused on the next generation
enterprise software startups.
Finally, he started teaching atStanford University in 1982, and
launched the university's firstclass on cloud computing. For
those of you that listened toour podcast last week, this is
(01:30):
where he met his now greatfriend, Dr. Anthony Chang. And
where the shoots have an ideawere formed to begin a project
to connect all healthcaremachines in all children's
hospitals in the world, which,like the consumer internet, may
completely change children'shealth care on the planet.
Today, we discuss this story andunderstand more about the
opportunity that Timothy and histeam are trying to uncover. We
(01:52):
really hope you enjoyed thisconversation. Timothy, Hi, thank
you so much for joining us onthe not mini adults podcast.
Timothy (02:02):
Thanks for inviting me, David.
David (02:04):
It's an absolute pleasure. I think the story that
we're about to hear and theconversation that we've that
we've gotten the topic is prettydifferent to what we've had
previously. So we're prettyexcited to share your vision,
what you're doing and know theinitiative that you're working
on. And it's all going to berevolving around clouds, right.
But I think to begin with, whenHannah and I were talking about
this conversation, the questionthat came up was, how did you
(02:27):
get into paediatrics? So can youtalk a little bit about your
kind of journey and know how yougot into doing what you're
doing?
Timothy (02:36):
Yeah. Be glad to. So my background, which is probably
unlike a lot of your guest isall coming out of IT and tech.
So I actually came to SiliconValley in the early 80s. To work
for one of the first KleinerPerkins startups, called tandem
computers, fundamentally havenever left the area, ended up
(02:57):
being the president, the firstpresident of Oracle's cloud
business. Actually, when I leftthere, I ended up starting the
first class on cloud computingat Stanford. And that is
actually the origin of how Icame to be involved or to
understand paediatric healthcare. My class and this is a
(03:18):
pre COVID conversation. So hasabout 100 people in it. About 90
of them are local and about 10are remote. So one of the remote
students reaches out and says,I'm a little old school. I like
to meet the professor. I went,Okay, cool. So we arranged to
have breakfast at Johnny's cafeon California Avenue. And I'm
sitting there and this guy walksin, I'm looking at him going, he
(03:40):
doesn't look like a regularstudent, right? sits down, come
to learn. He has an MD and MPH,an MBA, and he's chief of
paediatric cardiology atChildren's Hospital, Orange
County. And so I'm looking athim going well, why are you
talking to me? like what can Ihelp you with? So he said, Well,
actually, I think AI, Big Data,Cloud computing and medicine
(04:02):
should meet. And so I've comeback to Stanford to earn a
master's degree inbioinformatics. I love telling
this part of the story it takesAnthony three and a half years
to finish a two year programmebecause number one, he has no
idea how to code. Number two, hestill has a real job as chief of
(04:22):
paediatric cardiology. Andnumber three, he has decided
that in his 50's, to adopt an 18month old and a six month old as
a bachelor, I love tellingAnthony story because like it's
never too late for any of us.
Right. So Anthony is really myintroduction into the world of
paediatric medicine. And themore I got into it, the more I
(04:47):
was like, boy, we can help, wemeaning us in tech. We like to
talk about technology for good,but a lot of what we do is make
it easier to buy stuff and sellstuff. And I think this mission
really is a mission where we cantake our brains, our resources,
our connections, and really makea material difference in
(05:10):
Children's Healthcare.
David (05:13):
So for those that haven't listened to our previous episode
of the podcast, you're talkingabout Anthony Chang, who, who we
had on our previous episode.
He's, as you said, everythingthat you described, but also now
a data scientist, as a result ofis working with you and taking
your course. And, you know, wetalked a lot about AI and the
opportunities there and, and Ireally wanted to once we had
(05:33):
been introduced, really kind ofbring that in and kind of
continue the story and continuethat. Yeah, or just just the
story of what you know, what webegan with Anthony, and how
you've moved it forward. And Iguess there's a bit of kindred
spirits here. Because, you know,Hannah and I have been in
technology, all of our careersas well. And we strongly see the
opportunity for, you know, newtechnologies and bringing new
(05:54):
initiatives and innovation intopaediatrics. And that's exactly
what we're trying to do from acharity perspective is bringing
the future of healthcare tochildren. So continuing your
story you've touched on andwe're going to talk about
clouds, but you've started aninitiative with Antony's
support, looking at building apaediatric cloud. But before we
go into that, I think just forpotentially some of our
(06:17):
listeners, maybe you could justgive us a quick overview of what
cloud computing or what cloudis, you know, in terms of
technology, and how that couldever be relevant to healthcare.
Hannah (06:27):
And I'm so sorry, because I am going to chip in
and make it a double prongedquestion, because it was really
interesting to have the answerto David's question of cloud,
but also, why cloud? Why did youstart there, you had the meeting
of minds with yourself andAnthony, but then there was many
areas that the two of you couldhave hit on together to address?
Timothy (06:46):
Yeah, so maybe I'll kind of go down how we ended up
thinking about cloud being ananswer to the question. And
let's start with the question.
And then obviously, per David'scomment, I'll try to give a
brief tutorial on what cloudcomputing is. So what I started
to learn by virtue of beinginvolved with the community, was
just how I'll call it primitivethings are, so today, in fact,
(07:09):
this happened just last year, akid went undiagnosed with brain
cancer for over a month inSouthern California, because it
was too difficult to move an MRIimage through two hospitals and
one clinic. And I think a lot ofpeople in the world in the
paediatric world know this.
(07:32):
They're still using CD ROMs,which like, when's the last time
you saw one of those. And ofcourse, the CD ROM has a little
sticky note on it with thepassword. So guess how much
security that is? So we learnedthat I learned that story.
There's a story that Anthonytells, a little kid in Myanmar,
who he's providing consultingfor, and she's on the operating
(07:55):
room table, they want to sharethe echo with him in real time.
They can't do it. The kid dieson the operating room table.
Later on, he gets to see theecho says, Oh, I could have told
him exactly what was going on.
And then you start looking atthis globally, one of my old
friends is actually the CEO orthe former CEO of Save the
Children. And I was describingwhat we were looking at, she
(08:17):
says, Well, you know, pneumoniais the number one killer of kids
in Africa. And it's not becausethere's not low cost scanner
technology. There's nobody toread the scan. And in fact,
about a month ago, I'm spendingtime with leading paediatrician
in Rwanda. And I come todiscover there is a single one
paediatric cardiologist in theentire country, which then leads
(08:40):
obviously in the overlap withAnthony and AI, like, well
shoot, what if we could take thebest brains from Stanford,
children's Texas children,whatever, and put it in a
computer, so that you couldactually be able to diagnose
pneumonia or COVID, or whatevernumber of circumstances, when
you look at all of that, andsay, Well, what is the root
(09:04):
cause of everything from you'restill using MRI scan, you know,
CD ROMs, etc. And oh, AI isstill difficult, because and
everybody that's in the world ofAI knows this already to be
true, and particularly AI inmedicine is there's not enough
data to drive these deeplearning networks. And if you
(09:26):
don't have enough data, you'renever going to get very accurate
algorithms at all. So you lookacross this. You go well, what's
the big issue? The big issue isthat there's not data. I mean,
there's 1000s of CT scanners outthere, right. Hundreds of
ultrasounds, Gene sequencers butnothing is connected,
(09:47):
fundamentally. And so, I like tosay in our world, we cracked
through the connection problemback in 1994. We connected a
million computing machines, andthat was the beginning of the
consumer internet. And that'sthe beginning of eBay and
Netscape. And, and we all knownow at this stage, how much
(10:09):
that's completely transformedour consumer lives. And also, by
this point in time, I hadstarted to understand, there are
about 2000 healthcare machinesin every Children's Hospital in
the world. And there are about500 children's hospitals, which
turns out, that's about amillion healthcare machines. So
(10:30):
a year ago, we launched thisproject to connect all million
healthcare machines and all thechildren's hospitals in the
world, and create a digitalinfrastructure that could
transform children's healthcare. That's the problem, the
challenge we saw, kind of ourour thinking about it obviously,
(10:51):
is on a planet wide basis, whichleads to Okay, sounds like a
great vision, how the hell areyou going to do this? So if you
look at it go, Well, if I'm agene sequencer, a blood
analyzer, whatever, and I wantto get to that data, I need a
computer, we're going to call itan edge server to talk to that
(11:13):
machine. I'm also going to needa computer to talk out, for
instance, real time stream andecho as an example. So that
computer we call an edge server.
And so what we are building orwhat we have built is an edge
cloud. Okay, why do I call it anedge cloud? So I'll get to
(11:37):
David's question. So if you tookmy class, the very first day, I
tried to explain cloudcomputing, particularly what I
call compute and storage, cloudcomputing. And if you speak,
Amazon speak, I'm talking aboutwhat they call a EC2 and S3. So
what does Amazon do? Right?
Well, Amazon says, I'm going togo buy a million servers, a
(11:59):
million computers, I'm going tomanage the performance,
availability and security ofthose computers, I'm going to
deliver him in an op x model,you don't have to purchase up
front, you can pay me everymonth. And I'm going to put
those computers in 10 datacentres in the world. And what
I've just also explained to youas well, that's Microsoft Azure,
or whether that's Google Cloud,etc, that abstraction applies
(12:22):
across the board. Okay, so thatis cloud computing, I am
delivering compute and storageas a service by managing those
servers, and changing thebusiness model to saying yes,
you can buy it, frankly, by thehour by the day by the month,
what are we doing? So our edgecloud is in many ways the same
(12:47):
thing. So we're going well,we're going to buy a million
computers, we're going to managethe performance, availability
and security of those computers,we're going to deliver them an
op x model, the only thingdifferent is we're going to put
them in a million data centres,meaning right next to connected
to the ultrasound, the genesequencer, the blood analyzer,
(13:07):
etc. And the second thing we'regoing to do differently is we're
going to fill them with data. Sowe're going to fill them both
with what we call machine data.
So that could be things like,what's the serial number of the
machine? Where's it located?
What's the laser power level onthe gene sequencer, and the
other kind of data as we callit, gnomic data, so that's the
(13:27):
data the machine would generate.
So a blood analysis a genomesequence, an echocardiogram. And
based on that allowapplications, what we call edge
applications, to be able to doany number of different things
with that data, withpermissions, obviously. So it
(13:49):
could be things which are, I'mgoing to take that data, and I'm
going to stream it to a securecell service that looks like an
Instagram, we're actuallyworking with a company called
Figure One is that to aggregateit in a centre cloud to do
centre cloud, what we're gonnacall centre cloud learning, or
(14:11):
is that and we're just in themiddle of this is it to use it
to develop federated learningapplications, all of these
become possible as to what edgeapplications do. And therefore
we're only limited by people'simagination of what becomes
possible in all this.
David (14:32):
So I'm going to try and slightly simplify what I think
you've just said just forprobably from my own benefit,
but certainly for maybe some ofour listeners, but essentially
your aim is to take paediatricdata which if you took it from
one source or one hospitalwouldn't really be enough to
actually do anything useful withpotentially. Amalgamate that
(14:54):
together so that anyone anywherecan take advantage of that data.
In order to try and solve someof the, you know, the world's
biggest, biggest problems inchild health by putting it up
onto a central system, whichwhich people can gain access to.
Timothy (15:11):
we will enable just to be clear, we will enable that
what you just said, which isaggregation in a centre cloud,
we will also enable that itnever has to be aggregated. This
is an interesting thing. So Isaid the number 2000. So, if you
look, and we just ran anexperiment at Lucile Packard, so
(15:31):
if you imagine Lucile Packard abuilding, which now has 2000
servers in it, right, that allhave an interesting amount of
resource memory, GPU, CPU, allthat good stuff. And now
connected with a highperformance private 5g network,
you could imagine that what Ijust said as I just move the
(15:54):
cloud down into Lucile Packard'sbuilding, and now all the
computing that you want to do,from even from a learning
perspective, could actually alloccur locally. I mean, this
really stretches I think, wherewe are in technology. But that
is entirely possible, just tomake a point of it. It doesn't
have to be just about learning.
This could be as simple as theexample I gave you, which is,
(16:17):
there's a kid in Myanmar, I wantto share the echo image with
Anthony in Orange County, that'sall I want to do. This
infrastructure makes that easyto do. So it can be from the
very, let's call it far reachingalmost science fiction, down to
something very simple. I justwant to share a movie, like I
(16:37):
share with my friends onInstagram. Does that makes
sense?
David (16:43):
Yeah, absolutely. I think the vision is incredible. I
guess the next question is, youknow, people that know this kind
of computing, will some of thequestions will be security,
that's a fundamental thing thatpeople will be asking about. So
let's maybe go there first. Sode identified data security or
this element? How are you goingto kind of provide that and how
(17:05):
you're going to manage that?
Timothy (17:06):
Yeah. So maybe, obviously, when we launched
this, we knew security andprivacy was had to be designed
in from the beginning, weweren't going to kind of try to
add it at the end. So from anengineering perspective, we
started there, we've actuallyengineered 28 different security
features, and compare them towhat is being done at AWS, and
(17:27):
actually, we're doing more. Partof that has to do with the fact
that we've got problems that aremore difficult to solve. So in
their world, it's kind ofdifficult to walk in the
computer room and walk away witha server in our world, you could
actually walk away with aserver. So we have to engineer a
bunch more protectionmechanisms. The other part is
privacy. Kind of fortunate. Oneof my former students is a
(17:51):
privacy lawyer with 12 years ofexperience in privacy law,
particularly the GDPR. Soeverything we've done from a
privacy point of view has beenarchitected from day one to be
around GDPR, because we knew wewanted to go global. And we see
it as a superset of what HIPAcompliance is in the US. So
(18:13):
security and privacy have been aday one objective. We think we
built a world class system here.
And as we've been going out andtalking to the various hospitals
as we deploy, we're invitingthem to go, Hey, if we missed
something, we want to know aboutit. Because this is not
something we're trying to shyaway from. We think it's an
important bedrock issue. And wewant to bring the best of
(18:36):
technology to it.
Hannah (18:39):
While I'm having that side of the capability nailed
then gives you the opportunityto go global, as you just
referred to, as you were talkingin the conversation so far,
there's a topic that's come uptime and time again, in
conversations we've had on thenot mini adults podcast, and
that's been one of equality. Soyou talked about, first of all
(19:00):
you talked about, in my words,it would be accelerating, or
reducing the time it takes todiagnose an unwell child. And
that's been at the core of ourambition for our small charity.
That was one of our firststepping stones that we were
interested in, because we couldsee the promise that technology
had. But then what I've learnedsubsequently has been more
(19:23):
around what big problem wouldyou love to resolve in
healthcare time and time again,we've heard just that there are
many complex, interrelatedfactors that lead to a very
unequal provision of health. Andso an example of a first world
country versus a third worldcountry is and is an obvious
one. But to what extent were youdriven by this sort of global
(19:47):
agenda of getting a higher levelof health care to children,
regardless of where they wereliving? Was it more about speed
or diagnosis? Do you care moreabout the global agenda? What
how did those factorsinterrelate? And what's your
thoughts on that? equality ormore quickly inequality?
Timothy (20:03):
Yeah, and you know, we're all are seeing inequality
in spades right with COVID. So,no, I, I think it's because back
to what can we as technologistsdo, that can materially impact
the world has been the agenda?
And I told you when am I, one ofmy old friends is CEO of save
the children talked aboutpneumonia? You don't really have
(20:24):
to. I mean, by the way, in myopinion, you don't have to go as
far away as Africa to seeinequality. I mean, rural
California versus MetroCalifornia level of health care
is not the same. And why is itnot the same? Well, it's because
it's a manual system, based onsilos of humans and data. So if
I've got the best and brightestin Palo Alto, well, I'm lucky,
(20:48):
right? I live here. You know,you live in you know, in
Northern California, the inUkiah, whatever, you're not
going to see that you don't haveaccess to that. And I think we
know that technology can evenaccess to information, we see it
all the time. That is theconsumer internet. I mean,
that's the other thing that Ithink a lot of times, because of
(21:11):
where healthcare has come from,people have thought about
medical devices, and pharma anddrugs and all that sort of
stuff, which all have enormousmanufacturing costs and cost to
deliver. Because we all live inthe world of software, I'm going
software, the cost ofmanufacture is zero, it's your
brain power, cost to distributeis zero. So if we can enable a
(21:33):
platform and infrastructure,which allows the best and
brightest to use their brains tocreate diagnostics,
recommendation, any number ofdifferent things, we know the
cost to build is zero, we needto know the cost to deliver is
zero, we could now changechildren's health care. I don't
(21:54):
care whether you want to talkabout rural California, rural
England, or Africa, we have thecapability. It's all sitting
here in front of us. So whydon't we do it?
Hannah (22:06):
Given that the usual and oftentimes an obstacle might be
cost? And that's not the issuehere. What obstacles do you see
in front of you? And what otherhelp might you need?
Timothy (22:17):
Well, you know, I think the obstacles are not
technological. The obstacles areboth the obstacles of change of
anything, right. And as all ofus who worked in, in IT know,
right, transformation, digitaltransformation is not a trivial
thing. I don't care whatindustry you're in healthcare
(22:39):
doesn't make it any easier.
Let's say it that way. Right. Sothe obstacles of change, which
have social things in it, youknow, organisational, etc, are
all you know, and obviously,there are challenges around
security and privacy. But I haveto tell you, I feel like all of
that we have the technology toaddress, I think what we need is
(23:00):
what we all need. When you dodigital transformation, you need
the champions, right. You needthose early adopters to go
"Yeah, I see your vision. AndI'm gonna rally my troops over
to here". Anthony is the PiedPiper have this, by the way,
just to say it. I mean, he's outthere way in front of most of
(23:21):
them, talking about artificialintelligence and what it could
do in medicine. But he's anoutlier. I mean he's a real
outlier. And I think that isreally our biggest obstacle. And
just to be nice to the medicalclinical community, I tell my
world, you know, our world, Igo, you know, if I'm a
(23:41):
clinician, I live in the hereand now I have to live in the
here and now. And now this isall I have, I can't live a year
from now or five years from now,it makes no sense. And yet, what
we all live in and what we'reused to, and what we talk about,
is we talk about where thingscould be. And I think that
disconnect between how peoplethink, right is is there now
(24:05):
we're not you know, we arefinding and, you know, between
Anthony and many other people ata lot of these children's
hospitals, there are the earlyadopters, the Vanguard's the
people who are going, Yeah, wereally need to start doing this
differently. And I think ourchallenge is how do we rally
those troops, bring themtogether? galvanise this
(24:27):
mission? Whenever we meet withpeople, I always invite them to
be part of our mission. I don'tsee it as an us them question.
Yes, we're developingtechnology, etc. But it's an all
of us thing. And I think that ifwe get the right people,
galvanise them, right, motivate,help them right, we'll get here,
(24:49):
the technology part, I'll justkeep repeating. This is not
gonna be that hard. It's thenext step which will be hard.
Hannah (25:00):
Well, and the disconnect piece is interesting. Because I
imagine you're thinking aboutyour how you make it easier to
get from where we are today tothe future. So you're showing
people the way perhaps.
Timothy (25:14):
Both showing and helping, I think the beauty of
the software business is wedon't go back to it doesn't cost
anything. So every step of theway, if we can build it, I was
just talking to a systemintegrator yesterday, who wants
to be a part of this. And Isaid, I have to tell, you know,
when I go talk to people infinancial services, I can kind
(25:37):
of say, Well, let me you want toimagine the future, this is what
it could look like, I guess theyusually can do that, because
they've been living in the worldof software for a long time now,
because there is no physicalmeaning to the world they live
in. But in medicine, that'stotally not the case. And I
said, I think it's impinnchentto begin to show them this, so
(25:59):
that they can begin to graspwhat it is, and not leave it on
a PowerPoint slide or, a speechor whatever. And I think by
doing that, and you know, again,the world of software, we can
iterate at the speed of thought,I mean, you want to build it new
again, tomorrow, no problem, wecan do that, right? And get that
(26:19):
and bring people along and showthem the power of doing this. I
do believe they all realise thatthings got to change. It's just
how does that change happen?
Where does it happen? who leadsthat change? I think those are
going to be the centralquestions. And I think that's
the work, you guys are doinggreat. I mean, we need more
(26:40):
people to want to change thisthing. And then to see how
change can happen. And I thinkin that I've already seen this,
I do have the privilege ofteaching at Stanford. So I'm on
all the faculty distributionlists. So this, this quarter,
there are three AI classes beingtaught. And so I thought, just
(27:00):
for the fun of it, I'm justgoing to post into one of them.
Because they were all goingwell, we need projects we need
projects to work on. So I justposted in one of them, basically
not much information. I justsaid, How about a project to do
AI in paediatric medicine? Youknow, within 36 hours, I had
three teams of three peoplereach out, what can we do? What
(27:23):
problems should we work on? Howcan we help? You know, I think
the will is there in thecommunity in the larger
community. And I think bybuilding a platform, a way to
begin to do something aboutthis, then we can actualize that
we can take, you know, smartkids who want to work on ,who
are maybe brilliant in federatedlearning, and marrying them up
(27:46):
with clinicians who have thevision to say, this is what we
could do with it. And they Imean, I'll just give her real
credit that we've been workingwith a doctor at Children's of
Atlanta, Rito , I talked to herand I said, Well, hey Rito what
would if we had all the data inthe world? And she said,
(28:08):
paediatric cardiology paediatricecho expert and I said if we had
all the data in the world whatwould you solve? So she goes
1,2,3,4,5 .Okay, so. So now I'mlooking at 1,2,3,4,5 going, you
know, some of these words like aand and the ,I understand. But
most of these words like, whatdoes this mean? So I invited her
(28:31):
we have a team meeting everyMonday and Thursday at two
o'clock. So I said, Hey, Ritowill you come and do a, I called
it a micro MD degree? I said,you know, words, like, what is
cardiomyopathy mean? And shecame, I was so impressed, she
(28:52):
built a presentation, just forus, it was like, let me explain
this to you. Let me explain whyyou could understand it, what
the challenges are, and tons ofvideos so that we could go Oh, I
see what she means by that. Andso I think that's the other half
of this problem, which is, youknow, we need to bring the, you
(29:15):
know, ML federated learning, youknow, AI people, and we all know
how to talk tech head, but wealso need to find a way to learn
to talk to them. And some ofthat, I think, is they need to
educate us a little bit onwhat's the vocabulary that we
can start to use. And I think inthat merger of talent of the
(29:35):
domain, expertise of a Rito andthey, you know, federated
learning expertise is coming outof Stanford or, I mean, a lot of
this work actually is occurringat Oxford, by the way, right.
And we bring these communitiestogether on top of a platform
that lets them build it, youknow, we're not that far away.
David (29:56):
We're doing a lot of you know, given that we're hopefully
people are listening Listen tothis, but we're doing a lot of
nodding along with everythingthat you're saying. And I think,
you know, that's that's probablynot evident if you're listening.
So in total agreement, but oneof the things that really stood
out for me, what you just saidis, if you actually raise the
problem, and you start to, youknow, just put it out there,
(30:18):
what can you do to help us inpaediatrics, people want to do
things they want to help? Youknow, it's a it's an evocative
subject. But I think then, aswith anything you've got, you
know, as soon as it kind of goesout of the mind's eye, then you
get back to normal and peoplejust kind of carry on. So it's a
case of continuing to have it inin the mind's eye. I guess the
next question is, is the how,how are you going to, you know,
(30:41):
what do you need to do in orderto move this forward? And you've
already touched on a little bit,but you've, you've got your
first implementation you'reworking at, funnily enough,
Stanford children's, but can youjust talk us to a little bit
about what you've done already?
And then you know, what you'rehoping to do in the future? And,
and you've got a, you've got akind of vision of a domino
effect, I guess,
Timothy (31:00):
you know, our team is all people that come out of
enterprise software. So we kindof think in terms of well, how
do you get to the customer? Andhow do you grow the business, so
we're not academics. So we, weestablished a beachhead, fairly,
obviously, at Stanford, which,by the way, it wasn't even with
all the connection networks, andeverything, still was not the
(31:22):
simplest thing in the world. Sowe understand the challenges of
their organisational structures,and kind of the established ways
of thinking about things, evenat a leading institution like
Stanford, right, so. So that'sgreat, we learned a lot, we went
in, just to tell the story, wewent and installed the edge
server, it took two and a halfhours to do the installation,
(31:46):
which acts on the one hand, andit's not so bad, because we
actually have never seen anultrasound, we had no way to
test until we showed up in thein the building. But the flip
side is two and a half hours,you're going to deploy a million
of these that we just can't dothat. So we've already done an
iteration of the software andactually rereleased it onto the
(32:07):
edge server at Stanford, whichwe believe could take this
provisioning time down to like20 minutes. So we're improving
all along the way.
David (32:16):
So just quickly, what's an edge server for those that
might not know what is it youknow?
Timothy (32:22):
So the the edge server, the edge server is a computer
about the size of your hand,we're actually looking at some
that are two inches by twoinches, but the ones we're
working with right now, by thesize, your hand, which connect
to in the case of Stanfordconnect to the Philips epic
seven ultrasound, so that everyultrasound that is done that is
(32:43):
shipped to the PAC System isshipped to the edge server. And
then the edge server has twoapplications running on it. One
which looks at the gnomic data,the question you asked earlier
David, identifies that at theedge, and then actually, for
demonstration purposes, puts itinto a box account, because
there's a nice daikon viewersitting there. The other
(33:06):
application is with a thirdparty ISV named Unitsa, which
has been managing smart devicesand for people like apple and
carvanha, and whatnot. So I'veknown the founders for a long
time, we said, well, we couldbring all that technology to
managing healthcare machines,the maintenance of health care
(33:27):
machines, utilisation of healthcare machines, etc. So that's
the second application runs onit, which communicates with them
that's the which actually is anapplication that runs at AWS. So
that edge server is that device,which those applications run on,
which communicates with theexternal world. And with the
Philips, in this case, thePhilips ultrasound. So we
(33:49):
completed that it passed all thetests. Actually, right now,
specifically, we're stillworking. There's an integration,
which we built to get the syslogdata off the Philips machine,
which is all about machine data,and that we're going through
testing right now. So the nextbig step is we said, You know, I
think we know how to scaleinside the building. How do we
(34:11):
scale across institutions. Andso the next big thing as we've
called this, the vanguardproject, this is eight
hospitals, seven more. One ofthem is obviously Stanford,
seven hospitals, where we inessence, want to replicate what
happened at Stanford, to ensurethat we have built a scalable,
(34:32):
secure environment. So we are inconversations with about 15
different hospitals out there. Alot of names that you would
know, just to say it justyesterday, but being gay Sue,
which is one of the largestchildren's hospitals in Europe,
has agreed to be one of thevanguard hospitals. So our plan
(34:53):
is to identify the eight we'recovering. One of the reasons its
eight is we're covering all thecosts of doing this. So we plan
to identify the eight by the endof May, and start installations
in June. So that's our next bigstep is moving out to eight
different sites. Once we've donethat, I'm pretty sure not that
(35:15):
we know everything. But we knowa lot about what things we need
to work on for scalability,security and performance. And
we'll do another iteration testthat but I think, you know, in
this year, we will know what ittakes to deploy a million
servers.
David (35:32):
I mean, as we've kind of already said, the vision is
incredible. What you kind ofneed to do to realise that
vision, given that it's amillion servers in every
continent on earth, and the restof it seems, seems to be, you
know, quite quite a challengeand what have you. So what, what
do you envisage the futurelooking like? And when?
Timothy (35:51):
So, you know, some people ask me a question, why
children's medicine, right? Imean, doesn't adult medicine
have the same problem? Like, Oh,well, yeah, adult medicine does
have the same problem. But theidea that we would have any
prayer of doing it in everyadult hospital, I have no idea
how you'd pull that off. Now,the nice side about children's
(36:13):
medicine, I made the pointearlier, there are about 500
children's hospitals in theworld. Well, you can guess what
that means is that about 25 ofthem, that are the leaders and
you guys have been in thisspace, I think we all could
write them down. We know whothey are. So basically, the
first objective is, let's getthe 25, of which, by the way,
(36:36):
eight, are going to be in thiscategory already. Because if I
get the 25, the other 475, well,will fall in that's not going to
be that difficult. And I that'swhy this market is somewhat
unique, I think and the prayerthat we could pull this off,
because it's this type ofmarket, I think is very high. So
(36:59):
the step after the eight, Ithink will really involve we're
already having a ton ofconversations about this. There
are two angles to the next stepone is research. So what
research projects, which youcould have never done before,
are now possible, given thisinfrastructure. And so we're
(37:20):
having conversation. In fact, Ijust had one this morning, a
children's national who's beendoing a lot of work in federated
learning, is it going to befederated learning around CT
scan data for COVID prediction?
Is that what the thing is right,or what is it going to be? And
this obviously lines up with thequestion of what becomes
fundable? The one thing I say topeople is, I'm not sure how AI
(37:43):
and adult medicine gets funded.
But I'm pretty sure there's noeasy answer to AI and paediatric
medicine. No venture capitalistis going to invest in that
market. It's just not bigenough. So it's gonna have to be
a combination of, you know,foundations, government etc
(38:05):
money that's going to have tocome in just to pick on him for
a second, we did an analysis ofNIH funding. So NIH spends about
$1.8 billion a year inpaediatric medicine $1.8
billion. We tried to identifylike, how much of that is spent
(38:25):
on AI? 15 million, maybe? Andwe're stretching, right? So
there's an education cycle.
There's a lot to bring theresearch community together. And
they have responsibility in thisas well as like, well, what are
the big clinical issues? What'sthe proposal, we're working with
a team at Stanford, you couldguess in a project called
(38:48):
chamber to do aggregatedpaediatric echo data, and
looking at the potential of thatbeing funded by a number of
different sources. So I thinkthere's one whole angle, which
is research, predominantlyaround AI, which by the way, and
I tell their community this. Youknow, we have the chance in
paediatric medicine to leapfrogadult medicine, because they
(39:10):
have all the same issues. Butyou know, how are they going to
get aggregate data? How are theygoing to do federated learning?
I mean, you know, if thiscommunity gets together, we have
the potential to leapfrog whatthey're doing. So I think that's
one whole angle that we aregoing to press forward with the
other whole angle is productionapplications. So you guys come
out of my world. So we're justtalking about ISVs independent
(39:34):
software vendors, who because ofthe technology we have built,
will enable them to do thingsthey have never done before. So
just a simple example, there's acompany out there called
Nautilus amusingly a companybuilt by the guys who built the
CD ROM exchange software. Andthey've now moved their
application into the cloud. Andso we're saying hey, let's build
(39:56):
a Nautilus edge application,which if childrens of Atlanta
decides to give it permission toaccess the echo or the CT scan,
you can now do image exchangeand annotation using the centre
cloud application at Nautilus.
(40:17):
And then I mentioned as clinicalapplication to help the clinical
engineer who today is managing.
I mean, at Stanford, there's20,000 assets in the building.
And they're managing in thedisconnected state, they don't
know where the machine is, orwhether it's being used. So you
know, if you have too manymachines, do you have too few
(40:37):
machines, no one knows. Sothey're going to be more, these
are just early kind of peoplethat we've been working with
ISVs, that we want them to usethe infrastructure to develop
new applications or extend theircurrent application. So those
are the two big directions.
David (40:57):
Typically, we could literally talk about this, given
that this is kind of our area ofinterest. Very conscious, you
know, of your time. I'm verygrateful for the time that
you've given us. One of thethings that stand out is, and
it's great that you know, thiskind of thing is happening in
pedes. First, and I think we canall agree that if it if it
(41:19):
works, which we all have fullfaith, it will, then the
opportunity to scale up toadults is there and hopefully it
has been, it makes it a loteasier from that point of view.
And that's kind of exactly theessence of what we're trying to
convey for, from not mini adultspodcast perspective, I've got a
couple of kind of closingquestions for you. One is in one
(41:41):
sentence, you've talked about ita lot, but I'm just conscious
that some people won'tunderstand necessarily tell us
what federated learning is inkind of one sentence or, or
maybe two.
Timothy (41:52):
I don't know that I'm good enough to do it in one
sentence, but I'll explain itthis way. In contrast to
federated learning, classiclearning is, I take data from
many different sources. And Ibring it into a central
location. And I try to train myalgorithm using all the data in
a central place. What federatedlearning says is you don't need
(42:16):
to do that you can leave it inplace, and only transmit model
parameters. So it's all beenarchitected. In fact, if there's
a great video that Andy Traskdoes, you watch this thing, it's
all been done for healthcare,because of privacy concerns. So
rather than transmit the dataanywhere, just transmit model
(42:39):
parameters, so do the learningat the edge is how we would say
it right. So that you never haveto let the data leave the third
floor of alder Hey, just to pickon him. Right?
David (42:52):
Yeah, no, perfect. Thank you. And if anyone so if there
is someone from a hospital thatmay not be in that eight that is
listening and is interested,what do they need to do?
Timothy (43:02):
Oh, reach out. We'd love to talk. Do you guys, you
put a email address down. Justshare the details. I just as a
comment, I'm on LinkedIn toreach out. And LinkedIn is
another easy answer to this. Weare a consortium of the willing.
So anybody willing, come ondown?
David (43:23):
Well, so we will certainly share all your details
on that on the notes, soeveryone can can hopefully get
in contact. Final question thatwe ask everybody that comes on
to the podcast, Timothy, whichis if you could solve one
problem within paediatrics, whatwould it be?
Timothy (43:39):
Well, it sounds self serving. But I think we are
solving the one problem,paediatrics. You know if we
could pull this off, I don'tactually think it's a if it's
when we pull this off. It willchange everything. I mean, the
evidence I have of this is justimagine our world before 1994.
And imagine the world today,it's so completely different,
(44:01):
because we live in the connectedstate. And they live today in
that fractured siloed world thatwe all used to live in back in
the 80s. And that level ofchange is possible with
technology. And the willing andI you know, like I said,
probably self serving, but Ithink this is the one thing.
David (44:23):
Thank you. I had to ask the question. I've failed to ask
it previously, or at least tryto guess what the answer is and
got it entirely wrong but Ithought you'd say that. I'm glad
you said that. Thank you so muchfor joining us. Thank you for
your time, I think, you know,it's been a unique and
fascinating conversation, and Ihope everyone will agree with
that. So really appreciate it.
Timothy (44:44):
I appreciate you guys taking the effort to do
something like this, you know,without getting communicating
without sharing ideas is we'renot going to make it very far.
So I'll equally say thank youguys for doing these podcasts.
And I love the name Not miniadult. It's great.
David (45:06):
Thank you so much to Timothy for joining us on the
podcast and we wish him everysuccess in his endeavour, we can
really see how connecting theworld's paediatric data can have
such a huge impact on the healthof our children and children for
generations to come. Next week,we're bringing the conversation
back to the UK, and we're joinedby Dr. Don Sharkey to discuss
(45:29):
innovations and technologieswithin neonatal care. We really
hope that you can join us thenplease do subscribe to the
podcast. And if you're enjoyingit, please do leave us a review
as well. We hope you'll join usagain next week.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
Las Culturistas with Matt Rogers and Bowen Yang
Ding dong! Join your culture consultants, Matt Rogers and Bowen Yang, on an unforgettable journey into the beating heart of CULTURE. Alongside sizzling special guests, they GET INTO the hottest pop-culture moments of the day and the formative cultural experiences that turned them into Culturistas. Produced by the Big Money Players Network and iHeartRadio.