Hello. Hello. Hi.
So, the next speaker of today
is joining us remotely via Zoom.
So you will see her come up in the screen very soon.
Her name is Alessandra Sala.
She’s the Senior Director of AI
and Data Science at Shutterstock,
and President of Women in AI.
Her talk is called, Blending Art and Science,
A human Approach Towards the Future of Creativity.
Alessandra will discuss about how to incorporate
inclusive design principles into AI technology.
So let’s welcome, Alessandra.
[audience clapping] Hello, everyone.
Hello. I hope you hear me.
Thank you for having me today.
And I have prepared for you
a few entertaining slide, I hope.
As I’m not with you in person,
I hope we can go through some visual element
to really get to the core
of what’s happening in the creative world,
which I believe is at the core of the discussion today.
So blending art and science,
and by science I really mean AI.
So, a couple of words about Shutterstock,
it’s not a new company.
We started in 2003,
and has evolved along the years,
acquiring and growing the content for creative purposes
of different kind,
of different style for different purposes.
And in the last few years, starting in 2021,
as I joined Shutterstock,
we have doubled down our attention to the world of AI.
Buying a few extra startup in the AI space,
and continuing to evolve the reach and the breadth
of digital content in terms of images, video, 3D,
editorial content, studio, and more.
In the last few years,
I come from a scientific background, before in academia,
then in research center in industry, and today about labs.
And we have worked in AI for many, many years now already.
However, within the last year, year and a half,
we have seen dramatic change in this field.
And I want to touch on a couple of trends
that have enabled us.
And obviously, I’m talking about Generative AI.
This amazing technology,
that in the world of images and creative content,
it allows you to represent anything that you can think.
Just writing a little prompt, and here you go,
amazing representation out of our imagination.
But as all of these seems surprising,
seems amazing, seems inspiring,
let’s also look into what’s happening underneath.
What I want to share
are a couple of specific technical aspects,
but really at a high level.
So, in the early part of 2000,
what we have done
is we were really working and using AI technology.
Think about of an engine of a Vespa, right?
In the last few years,
we are using engine of a Ferrari.
And how do we fuel those amazing engine with data?
Unfortunately, what has happened
is that the practice that has been used for company
that have shared those model open source,
was to grab everything available in internet
without any moderation.
Full of bias, full of violent content,
and nudity, non-safe for work,
everything that our humanity
has represented in internet so far,
and threw into those model, right?
Already few years ago,
Dr. Buolamwini showed that image recognition technology
used with pretty much the same approach
was discriminating against minority people of Black skin.
And now I know what you’re thinking.
Ale, yes, it was few years ago. [chuckles]
Now, everyone is talking about diversity and bias,
and this and that.
Yeah, I agree with you.
I thought we solved the problem as well.
But let me show you what those generative model
in the world of images that are open source out there,
do it today.
Again, the same bias.
So we didn’t solve the problem.
What we really want to focus on
is how can we embrace
ethical AI with a responsible approach
and rethink our data sourcing strategy.
So, the world is diverse.
The world is diverse in gender, ethnicity, age,
ability, feelings, emotion.
And that’s what the creative community,
the photographer, the artist,
are the best at capturing those unique model,
and see the world through diverse perspective.
And don’t we want
our best AI to inspire us along those dimension?
Not the one that I showed before.
So let’s play a very simple game with me.
On your hand, left and right,
help me to distinguish which of those two images
is generated from an AI system.
Count on the right to the one that you catch on the right,
and count on the left, the one that you catch on the left.
This is the first pair, left or right,
which is the AI-generated one?
Let me show you another example.
Oh, this is one of my favorite.
I’m originally from Amalfi, so I love the water.
So, beautiful water and some crops, left or right,
count on your hand.
Beautiful flower and an oil painting.
Count on your hand.
Let’s now see it.
Everyone was generated by our image generation tool.
So what am I trying to say?
Those tool, are today,
so good in representing
our reality, our imagination.
The things that we have fed into those system,
which may start to become difficult from a human brain
to differentiate which one is an AI creation
and which one is an authentic artist photography.
So, how do we manage this world, right?
How do we build technology that is fair for everyone?
Those were the question that we have been asking ourself
for the last couple of years.
So, November 9th, few days ago,
we did launch what we call the trust framework.
The trust framework is really moving
from ethical principle and guidelines,
to concrete action.
And today, I want to share with you
three key pillar behind the trust framework at Shutterstock.
How do we put artists at the core?
We have redesign…
The first one to redesign a business model
that compensate artists.
And the third one,
how do we enable transparency
to try to avoid to fool people
whether content is AI generated or not?
So let’s get started.
The first question is bringing artists at the core.
So, we have been constantly thinking about human art.
But what if art becomes a tool
in the end of our best creative mind
to bring us to a different level,
to inspire human creativity?
So we have worked with this artist, a Dutch artist,
Jeroen van der Most,
who has been working with data, algorithms,
and artificial intelligence for over 10 years,
and he’s an amazing person.
And his current thinking, it’s very much along the line,
How can AI inspire us to inspire our world,
and help us artists
to trigger and push our boundaries of creativities?
So we have embraced some of the learning
that we have done by talking with our artists,
by talking to our photographer,
and try to understand how they need this technology
to complement the creativity, the genuinity,
the inspiration that they put in the world.
And how do they make the full end-to-end creative process
a more pleasant process?
Because it’s amazing to take a picture,
but the editing process around it,
it’s much more a manual, and painful,
and takes a lot of time.
What if it existed a tool
that will help you to change the background,
reproduce just the perfect color
that you wanted to capture in that image,
create the perfect template for your design
with an intelligent tool that you can just type few words
and will complement the manual work,
that before, you had to go through and was so painful.
So the second pillar that I want to touch on
is the business model.
And on this one,
I want to pay a little bit of extra attention.
We have been the first one launching this idea
of What if we can compensate artist?
And I’m proud to say that this is a special industry.
The creative and media industry have demonstrated
that they intend to do the right thing.
are doing exactly the same things now, right?
The best representative possibly there with you today
from Adobe, Getty, and so on,
they’re compensating their artists as well.
And later on, I’ll show you how we have joined
some other activity led or started with our competitors.
So it’s an industry that is trying to learn
how we can do things differently,
and therefore, bringing the creativity, human creativity,
into this technological revolution.
So, what was happening
before we started talking about this new business model?
And what has happened
for a more than a decade or two on the web?
grabbing data from us in the social media,
or in our browsing activities,
or in our digital presence like maps,
building technology to automate or provide services,
selling back those services to client
with one single beneficiary.
So one person, entity, organization
accumulating all the benefit of this amazing technology.
What have we proposed?
A different approach starting with the content contributor,
They provide us licensed data.
But wait a second,
licensed data only if they want.
They have an option to opt out
so that we can respect their preferences.
Once we leverage,
only the best curated,
and metadata associated to those digital asset,
we create technology that is able to generate or to edit,
and to do amazing things in the creative space.
But we are not the only one
that share the benefit of this technology.
We share those benefit with all the contributors
that have decided to enable their creation
to build this technological innovation.
What does that mean?
Like, people holding stocks.
The more they have,
the more share the revenue
when those company pay dividend.
the idea of contributor
having their image there as their asset
that used in this engine,
generate revenue for them without the usual production cost.
And we have made it real.
We have announced two fund,
the Contributor Fund and the Creative Fund.
So the Contributor Fund is this fund mechanism
that the more this AI generative tool is used
and sell images,
the more those contributor receive compensation,
shared compensation, loyalty,
for that machine to create financial transaction.
The Creator Fund is a fund that provides
financial support and professional support
to usually excluded artists around the world.
So to allow different representation of our reality
through the eyes of of those
that are usually excluded voice.
And therefore, we continue to grow our content catalog
with diversity in mind.
So, now, going back to that idea
where we play the game of counting how many AI generated
or real photos were in those images that I showed,
here, it’s another one.
There are three generated from AI,
and one generated from a photo created from a photographer.
And it took me to three trial to identify the right one,
the top right corner, right?
How can you say that?
They are all amazing and beautiful.
But there is a mechanism.
And here, I want to go back to our industry,
a special industry.
We are sharing ideas and participating
to what we believe is the right thing.
Shutterstock is creating a new business model.
Our competitor believe that is a good idea.
They bring it to the market.
We see value in the content authenticity initiative,
which is a consortium that was initiated from Adobe.
And we have become member to support that initiative,
and also the member of the standardization group
behind the content authenticity initiative,
the Content Coalition for Authenticity and Provenance
that define an industry standard
to how to label and certify content,
so that user,
the news, the PVC,
anyone that wants to bring an image to their audience,
they can also bring those certificate along.
demonstrate whether it is an AI-generated content
or it’s a photographer at that event.
We want to be amazed by those real moment.
Look at here, Rihanna at the Super Bowl.
I was watching it.
And we want to see those amazing moment.
And we want to be able to acknowledge
through rigorous certificate
that this image was indeed taken from this photographer
at the Super Bowl
for that specific agency.
Or on the other side,
take an AI-generated image
and say, This was generated from Shutterstock
AI generative tool.
And certify, so that people that consume the content,
they understand the provenance and authenticity.
And we have seen recently, mainly for Gaza,
those that were using AI-generative tool to create content,
appropriate or less appropriate,
but the important thing is to make sure
that the consumer are aware
if they’re looking at a real picture
or an AI-generated picture.
So, bring into consumer and artist,
tools that are commercially safe.
It means that we don’t violate copyright
by grabbing anything available from Shutterstock
or available from internet.
We only use content that is properly licensed
to Shutterstock from the right holder.
And therefore, once we generate a new image,
we can attach a license,
and make that image legally compliant,
It goes through a content moderation process
that blocks any attempt to create violence,
nudity, non-safe content.
And therefore, when our enterprise customer,
they take specific action like Save or Add to Chart,
then we even send it through one of our human reviewer,
our expert of content,
to guarantee that that image is safe,
and therefore, we create indemnity production
for the company buying that image
for their commercial purposes.
And so, with that, I like to conclude,
and I don’t know if we’ll have any time for question today.
But otherwise, reach out to me, Alessandra Sala.
You can find me on LinkedIn.
This is an open discussion that we can only evolve
towards our North Star,
which is taking ethical lenses
and putting human at the center
of this technological revolution in the creative space,
into the media space, into our industry,
to allow this technology to enrich our human capabilities
in a way that respect the artist,
and also respect the consumer of this content,
giving them all the tool to identify
what kind of content they are consuming.
So, thank you for having me with you today.
Again, I’m sorry I’m not there with you in person.
I would have loved, to be honest.
But I hope my content has touched you
in a way that this discussion can be the beginning
of a future collaboration.
We have time for questions.
See if someone has a question for Alessandra.
Thank you, Alessandra.
[Alessandra] Thank you.