covering AWS re:Invent 2017, presented by AWS,

Intel, and our ecosystem of partners.

>> Hello everyone, welcome to a special

CUBE presentation here, live in Las Vegas for

Amazon Web Service's AWS re:Invent 2017.

This is theCUBE's fifth year here.

We've been watching the progression.

I'm John Furrier with Justin here as my co-host.

Our two next guests are Bob Rogers,

the chief data scientist at Intel,

and Julie Cardoa, who's the CEO of Thorn.

Great guests, showing some AI for good.

Intel, obviously, good citizen and great technology partner.

Welcome to theCUBE.

>> Thank you, thanks for having us!

>> So, I saw your talk you gave at

the Public Sector Breakfast this morning

here at re:Invent.

Packed house, fire marshal was kicking people out.

Really inspirational story.

Intel, we've talked at South by Southwest.

You guys are really doing a lot of AI for good.

That's the theme here.

You guys are doing incredible work.

>> Julie: Thank you.

>> Tell your story real quick.

>> Yeah, so Thorn is a nonprofit,

we started about five years ago,

and we are just specifically dedicated to build

new technologies to defend children form sexual abuse.

We were seeing that, as, you know,

new technologies emerge, there's new innovation out there,

how child sexual abuse was presenting itself

was changing dramatically.

So, everything from child sex trafficking online,

to the spread of child sexual abuse material,

livestreaming abuse, and there wasn't a concentrated effort

to put the best and brightest minds and technology together

to be a part of the solution,

and so that's what we do.

We build products to stop child abuse.

>> John: So you're a nonprofit?

>> Julie: Yep!

>> And you're in that public sector,

but you guys have made a great progress.

What's the story behind it?

How did you get to do so effective work

in such a short period of time as a nonprofit?

>> Well, I think there's a couple things to that.

One is, well, we learned a lot really quickly,

so what we're doing today is not what

we thought we would do five years ago.

We thought we were gonna talk to big companies,

and push them to do more,

and then we realized that we actually needed to be a hub.

We needed to build our own engineering teams,

we needed to build product,

and then bring in these companies to help us,

and to add to that,

but there had to be some there there,

and so we actually have evolved.

We're a nonprofit, but we are a product company.

We have two products used in 23 countries around the world,

stopping abuse every day.

And I think the other thing we learned is that

we really have to break down silos.

So, we didn't, in a lot of our development,

we didn't go the normal route of saying,

okay, well this is a law enforcement job,

so we're gonna go bid for a big government RFE.

We just went and built a tool and gave it

to a bunch of police officers and they said,

"Wow, this works really well,

"we're gonna keep using it."

And it kinda spread like wildfire.

>> And it's making a difference.

It's really been a great inspirational story.

Check out Thorn, amazing work,

real use case, in my mind, a testimonial for

how fast you can accelerate.

Congratulations.

Bob, I wanna get your take on this because

it's a data problem that, actually,

the technology's applying to a problem that

people have been trying to

crack the code on for a long time.

>> Yeah, well, it's interesting,

'cause the context is that we're really in this

era of AI explosion,

and AI is really computer systems that can do things

that only humans could do 10 years ago.

That's kind of my basic way of thinking about it,

so the problem of being able to recognize

when you're looking at two images of the same child,

which is the piece that we solved for Thorn,

actually, you know, is a great example of using

the current AI capabilities.

You start with the problem of,

if I show an algorithm two different images

of the same child, can it recognize that they're the same?

And you basically customize your training

to create a very specific capability.

Not a basic image recognition or facial recognition,

but a very specific capability

that's been trained with specific examples.

I was gonna say something about what

Julie was describing about their model.

Their model to create that there there

has been incredible because

it allows them to really focus our energy

into the right problems.

We have lots of technology,

we have lots of different ways of

doing AI and machine learning,

but when we get a focus on this is the data,

this is the exact problem we need to solve,

and this is the way it needs to work for law enforcement,

for National Center for Missing and Exploited Children.

It has really just turned the knob up to 11, so to speak.

>> I mean, this is an example where, I mean,

we always talk about how tech transformation

can make things go faster.

It's such an obvious problem.

I mean, it's almost everyone kinda looks away

because it's too hard.

So, I wanna ask you, how do people make this happen

for other areas for good?

So, for instance, you know,

what was the bottlenecks before?

What solved the problem, because, I mean,

you could really make a difference here.

You guys are.

>> Well, I think there's a couple things.

I think you hit on one, which is

this is a problem people turn away from.

It's really hard to look at.

And the other thing is is

there's not a lot of money to be made

in using advanced technology to find

missing and exploited children, right?

So, it did require the development

of a nonprofit that said, "We're gonna do this,

"and we're gonna fundraise to get it done."

But it also required us to look at it

from a technology angle, right?

I think a lot of times people look at social issues

from the impact angle, which we do,

but we said, "What if we looked at it

"from a different perspective?

"How can technology disrupt in this area?"

And then we made that the core of what we do,

and we partnered with all the other

amazing organizations that are doing the other work.

And I think, then, what Bob said was that

we created a hub where other experts could plug into,

and I think, in any other issue area that you're working on,

you can't just talk about it and convene people.

You actually have to build, and when you build,

you create a platform that others can add to,

and I think that is one of the core reasons why

we have seen so much progress,

is we started out convening and really realized

that wasn't gonna last very long,

and then we built, and once we started building, we scaled.

>> So, you got in the market quickly with something.

>> Yeah.

>> So, one of the issues with

any sort of criminal enterprise

is it tends to end up in a bit of an arms race,

so you've built this great technology but then

you've gotta keep one step ahead of the bad guys.

So, how are you actually doing that?

How are you continuing to invest in this and develop it

to make sure that you're always one step ahead?

>> So, I can address that on a couple of levels.

One is, you know, working with Thorn,

and I lead a program at Intel called

the Safer Children Program, where we work with

Thorn and also the National Center

for Missing and Exploited Children.

Those conversations bring in all of the tech giants,

and there's a little bit of sibling rivalry.

We're all trying to throw in our best tech.

So, I think we all wanna do as well as we can

for these partnerships.

The other thing is, just in very tactical terms,

working with Thorn, we've actually,

Thorn and with Microsoft, we've created a capability

to crowdsource more data

to help improve the accuracy

of these deep learning algorithms.

So, by getting critical mass around this problem,

we've actually now created enough visibility

that we're getting more and more data.

And as you said earlier, it's a data problem,

so if you have enough data,

you can actually create the models with the accuracy

and the capability that you need.

So, it starts to feed on itself.

>> Julie talked about the business logic,

how she attacked that.

That's really, 'cause I think one thing notable,

good use case, but from a tech perspective,

how does the cloud fit in with Intel specifically?

Because it really, the cloud is an enabler too.

>> Bob: Yeah, absolutely.

>> How's that all working with Intel?

And you go on about whole new territory

you guys are forging in here, it's awesome, but the cloud.

>> Right, so, for us, the cloud is

an incredible way for us to make our compute capability

available to anyone who needs to do computing,

especially in this data-driven algorithm era where

more and more machine learning, more and more AI,

more and more data-driven problems

are coming to the fore,

doing that work on the cloud and being able to

scale your work according to how much data

is coming in at any time,

it makes the cloud a really natural place for us.

And of course, Intel's hardware is a

core component of pretty much all the cloud

that you could connect to.

>> And the compute that you guys provide,

and Amazon adds to it, their cloud is impressive.

Now, I'd like to know what you guys

are gonna be talking about in your session.

You have a session here at re:Invent.

What's the title of the session, what's the agenda,

is it the same stuff here, what's gonna be talked about?

>> So, we're talking about life-changing AI applications,

and in specific we're gonna talk about,

at the end Julie will talk about

what Thorn has done with the child-finder and the AI

that we and Microsoft built for them.

We'll also, I'll start out by talking about

Intel's role broadly in the computing and AI space.

Intel really looks to take all of its different hardware,

and networking, and memory assets,

and make it possible for anybody to do the kinds of

artificial intelligence or machine learning

they need to do.

And then in the middle, there's a really cool

deployment on AWS sandwich that (something)

will talk about how they've taken the models

and really dialed them up in terms of

how fast you can go through this data,