00:00:02
december 2023 the New York Times becomes
00:00:05
the first major US media company to sue
00:00:08
Open AI for scraping millions of their
00:00:10
articles to train chat bots others soon
00:00:14
follow
00:00:15
suit in 2024 fast fashion giant Shiin is
00:00:20
accused of using AI to track and copy
00:00:22
designs from other sites giving them an
00:00:25
unfair edge it's a knockoff of my silk
00:00:29
lace cami to the tea it's exactly the
00:00:32
same soon other platforms like Meta
00:00:35
begin updating their terms allowing them
00:00:38
to create derivative AI content from
00:00:40
users posts and data adobe actually just
00:00:44
announced that they are stealing your
00:00:45
work and using it for a profit they're
00:00:48
spying on your computer every time you
00:00:50
open up Photoshop Illustrator After
00:00:52
Effects whatever app that you use
00:00:56
though hailed by technologists and users
00:00:58
for its innovative potential generative
00:01:00
AI also has many others worried
00:01:03
everything that I am my background my
00:01:06
history it's just all stripped away and
00:01:09
fed in through a machine now artists and
00:01:11
writers are saying "Wait a minute how
00:01:13
did you make these tools where did you
00:01:15
get all the material?" The stakes are
00:01:18
high with the Gen AI market exploding
00:01:21
across the world these individuals are
00:01:23
sounding the alarm it really is
00:01:26
something that can be used to reshape
00:01:28
our society and in the wrong hands uh
00:01:30
with the wrong intentions it could do
00:01:32
very bad
00:01:40
[Music]
00:01:43
things i photograph people for my work
00:01:46
primarily a lot of fashion and beauty
00:01:48
and celebrity work
00:01:53
yeah that's nice singaporean
00:01:55
photographer Jang Jingner's work is
00:01:57
recognized around the world she's become
00:02:00
known for her unique style that crosses
00:02:02
eastern and western art traditions
00:02:06
i kind of mix my Asian background
00:02:08
together with kind of painterly touch
00:02:10
and influences from anime manga and
00:02:12
fantasy art i think photography is an
00:02:16
art because it allows you creative
00:02:19
expression through imagination you know
00:02:21
willing something in your mind in to
00:02:24
existence i think that's very cool
00:02:28
in 2017 she created a compelling image
00:02:31
that would later spark an international
00:02:33
controversy this photo was taken in uh
00:02:36
November
00:02:38
2017 and photographed as part of a
00:02:41
series of photos for Hapas Bazar Vietnam
00:02:44
when we had a model's special feature
00:02:46
cover story
00:02:48
fast forward to 2022 jinga discovers
00:02:51
that a painting with an uncanny
00:02:53
resemblance to her photo is getting a
00:02:55
lot of attention
00:02:58
online i'm seeing that this particular
00:03:01
painting was receiving these accolades
00:03:03
and praise calling it original and
00:03:06
winning award and it was even shown on
00:03:09
newspapers and
00:03:12
TVs it felt like a slap on the face
00:03:16
awarded a prize at the prestigious
00:03:18
Luxembourg Bionali the painting
00:03:21
Churandot by artist Jeff Dishberg bears
00:03:24
striking similarities to Jingna's photo
00:03:28
this is my original photo and this is
00:03:30
the painting by Jeff in terms of
00:03:34
similarities well it's a flipped version
00:03:37
of my work you know the model's likeness
00:03:40
is quite close to how she looks and the
00:03:44
draping of the clothing the petals um
00:03:49
very closely replicated as well
00:03:53
winning a
00:03:54
€1,500 cash prize Dishberg puts his art
00:03:57
on sale for
00:03:59
€6,000 jingner decides to sue him for
00:04:02
copyright infringement in Luxembourg
00:04:05
i'm not pursuing this case for money is
00:04:08
really about trying to bring awareness
00:04:11
to the public that this is not okay to
00:04:14
do to artists and creators at first
00:04:17
Dishberg claims that his painting is an
00:04:19
example of the fair use principle using
00:04:22
an existing artwork to create a new one
00:04:26
the doctrine of fair use means you are
00:04:29
allowed to infringe on copyright in
00:04:32
certain contexts fair use is extremely
00:04:35
important to art and
00:04:40
journalism over the years the concept of
00:04:43
fair use has enabled the creation of all
00:04:45
kinds of new works in the arts and
00:04:47
entertainment industry from famous
00:04:50
paintings to musical parodies but a
00:04:53
prerequisite is that the new content
00:04:55
needs to transform the existing one in a
00:04:57
significant way
00:04:59
essentially when you look at whether
00:05:01
something is transformative you're
00:05:03
trying to figure out whether it makes
00:05:05
the original work different enough to
00:05:08
justify the copyright
00:05:10
infringement determining if fair use is
00:05:13
justified in art can be highly
00:05:15
controversial
00:05:16
in 2015 transformative artist Richard
00:05:20
Prince took the Instagram posts of
00:05:22
various social media users without
00:05:24
permission altered them with his own
00:05:26
comments and sold them in a gallery for
00:05:30
$100,000 each
00:05:33
a really important case that examined
00:05:36
this question of whether something was
00:05:38
transformative or not was just heard by
00:05:40
the Supreme Court
00:05:45
andy Warhol decided to paint Lynn
00:05:48
Goldsmith's photograph of Prince in a
00:05:51
magazine goldmith was arguing that it
00:05:53
wasn't transformative but also that it
00:05:55
caused market harm to her and to
00:05:57
photographers everywhere because if
00:05:59
you're allowed to just silkcreen a
00:06:01
version of a photograph how does that
00:06:04
not put photographers out of business
00:06:08
uh just lowering the power
00:06:13
sure sorry my career is made by me being
00:06:17
on the internet and I thought that would
00:06:19
be really good for younger generations
00:06:22
web generation artists and creators to
00:06:25
tell people hey you can't use my work
00:06:28
this way
00:06:30
jing Na's argument was that the painters
00:06:33
work was not transformative enough to
00:06:36
justify imitating her work
00:06:40
but copyright laws differ from country
00:06:42
to country here in Luxembourg Jingner
00:06:45
must also prove that her photograph is
00:06:47
an original creation
00:06:49
the important point is whether the
00:06:51
photograph was sufficiently original to
00:06:54
be protected by copyright and there the
00:06:56
judge felt that there was not enough
00:06:59
evidence to prove the originality of
00:07:02
Yinga's photos
00:07:05
in December 2022 the Luxembourg court
00:07:08
rules against Jingna
00:07:10
[Music]
00:07:13
i
00:07:13
was I was very devastated by by the
00:07:18
result the judge said that my work was
00:07:21
not original enough to receive copyright
00:07:23
protection and because my pose was not
00:07:26
original because it's a classic pose and
00:07:29
the thing is if a pose a singular thing
00:07:32
is what it takes to to receive copyright
00:07:34
protection then most of art in the world
00:07:39
are not going to receive copyright
00:07:40
protection and that does not seem right
00:07:45
undeterred Jingna appeals the ruling but
00:07:48
the case starts to attract more
00:07:50
attention
00:07:51
online and things take a darker
00:07:57
[Music]
00:08:04
turn my life has been on a hold i I
00:08:07
couldn't sleep normally for a long time
00:08:10
i would also receive continued
00:08:13
harassment for being a Chinese person
00:08:16
who is complaining about copyright about
00:08:19
me being a woman calling out men i will
00:08:23
also receive messages saying that I
00:08:26
should kill myself
00:08:28
soon i abandoned a lot of my web
00:08:31
presence online
00:08:34
while dealing with the fallout a new and
00:08:37
bigger threat emerges one that will
00:08:39
change Jinga's life and the face of
00:08:41
copyright
00:08:47
forever dolly 2 is a new AI system from
00:08:51
OpenAI that can take simple text
00:08:53
descriptions like a koala dunking a
00:08:55
basketball and turn them into
00:08:56
photorealistic images that have never
00:08:59
existed before
00:09:01
research organization OpenAI launches
00:09:04
Dari a generative AI tool that allows
00:09:07
people to produce images from
00:09:13
words other major companies like
00:09:15
Midjourney and Stable Diffusion soon
00:09:18
follow with similar AI programs enabling
00:09:21
the creation of any image in just mere
00:09:24
seconds
00:09:26
a lot of the major AI companies came out
00:09:29
with consumer AI tools that were trained
00:09:33
on the works of artists and
00:09:38
writers taking from publicly available
00:09:40
sources Genai learns from books websites
00:09:44
social media and more enabling it to
00:09:46
generate works that can often closely
00:09:49
resemble those of others leading to
00:09:51
accusations of copying suddenly anyone
00:09:55
could become an AI creator
00:10:01
i discovered that my work and my name
00:10:05
was being used in prompts for generative
00:10:07
AI pretty early on people use Gen AI to
00:10:11
make copies of this work and spam it a
00:10:14
100 times just to harass me right and
00:10:17
then now they're just putting in your
00:10:18
face to remind you well your work is
00:10:20
worth nothing i can copy it so many
00:10:22
times so easily
00:10:26
now artists and writers are saying "Wait
00:10:28
a minute how did you make these tools
00:10:30
where did you get all the material?"
00:10:33
Despite the harassment Jinga isn't
00:10:36
giving up for the sake of her future and
00:10:38
that of fellow creatives she's decided
00:10:41
to take the fight up a notch there were
00:10:44
also many many non-artists who don't
00:10:46
really understand how copyright works
00:10:49
and they would say that because uh she
00:10:52
posted her work on the internet it means
00:10:55
that she has given up copyright of her
00:10:56
work that misconception is so widespread
00:10:59
that we are seeing it with generative AI
00:11:01
companies right now so that has become
00:11:03
something I've been trying to fight and
00:11:05
raise more awareness for because it's
00:11:08
not right and it's not okay
00:11:25
embroiled in a copyright lawsuit over
00:11:28
her work renowned photographer Jung
00:11:30
Jinga is now facing another predicament
00:11:33
like many other artists around the world
00:11:35
her distinctive visual style is being
00:11:38
copied by people online using AI and it
00:11:42
all starts with a prompt
00:11:45
a prompt is oftentimes a sequence of
00:11:47
words that describes what you want to
00:11:48
see in the image they can say "I want
00:11:51
this to be cinematic i want this to be
00:11:52
high resolution i want this to not have
00:11:55
you know six fingers on her hand." It's
00:11:57
novel it's something that people have
00:11:59
never used before and so when you try
00:12:03
something which normally takes you 10
00:12:06
hours to do and is now done in 10
00:12:08
minutes it blows people's mind and so
00:12:11
it's always a factor of wow that you can
00:12:13
create a picture you can create a video
00:12:16
like that in seconds in minutes that's
00:12:23
groundbreaking boris L daxon from
00:12:25
Germany is among the new wave of artists
00:12:28
eager to show the world how to create
00:12:30
art using AI prompts once a traditional
00:12:33
photographer he now calls himself a
00:12:36
promptographer
00:12:38
for me promptography is working with AI
00:12:42
in collaboration i started two years ago
00:12:46
and I got sucked into it immediately it
00:12:48
was very addictive spending all nights
00:12:51
trying to create images with words also
00:12:54
completely new way of creating
00:12:57
images i'm going to show you how easy it
00:13:00
is to generate an image with text i'm
00:13:03
using midjourney as a platform a
00:13:08
chimpanzee and he should uh look like
00:13:11
me now AI is trying to come up with an
00:13:15
image that is related to the words I
00:13:18
used in the next step I just want to
00:13:20
improve the
00:13:22
classes something like mirrored ski
00:13:26
classes so here we are different
00:13:29
classes potentially I can steal them
00:13:32
from the internet i can photograph in
00:13:35
museums i can photograph books and
00:13:37
cataloges and the sad thing about it is
00:13:40
that style has no
00:13:43
copyright so right now there's nothing
00:13:46
you can do as an
00:13:49
artist with AI enabling anyone to create
00:13:52
endless versions of any image
00:13:54
traditional artists like Jingna don't
00:13:57
stand a chance the harm is on a personal
00:14:01
level how dehumanizing it has been to
00:14:04
see my name used in prompts tens of
00:14:06
thousands of times everything that I am
00:14:09
my background my history it's it's just
00:14:13
all stripped away and fed in through a
00:14:15
machine right there is no regard for my
00:14:19
copyright
00:14:23
jinga is not alone in her plight over
00:14:26
2,000 m away another artist is facing a
00:14:29
similar
00:14:30
battle for over a decade Kelly Mccernan
00:14:34
has built a reputation with a bold and
00:14:36
distinctive surrealistic style the work
00:14:40
that I create comes from a very very
00:14:42
deep place where I'm compelled to
00:14:46
express something whether it's trauma
00:14:49
whether it's love and I was able to put
00:14:52
my art on the internet and good faith
00:14:55
because I wanted to share it but I also
00:14:57
wanted to build community
00:15:00
like Jinga Kelly's art has been caught
00:15:02
up in the generative AI storm i believe
00:15:06
it was March
00:15:07
2022 when I began to see these images
00:15:12
being shared on Twitter with artists
00:15:15
being referenced but it didn't look like
00:15:17
their
00:15:25
work i didn't realize yet the
00:15:28
implications of all of this so
00:15:30
midjourney starts popping up as this
00:15:33
thing I keep hearing about in these
00:15:36
really strange
00:15:38
intricate images that don't really look
00:15:41
like a human made
00:15:46
them i do some more research and
00:15:49
discover that myself and many of my
00:15:53
peers ended up in it because we have
00:15:55
very distinct illustrative styles and at
00:15:58
that point that I think that my name had
00:16:00
been used thousands of times
00:16:05
already it was very
00:16:10
eerie i kind of felt like they were
00:16:12
unfinished sketches from my brain on the
00:16:16
screen there was a lot of shock and
00:16:18
wonder and an ominous feeling how oh
00:16:21
this is going to change everything
00:16:26
are you ready to learn about a free AI
00:16:28
art generator that can create this or
00:16:30
this or even this by feeding it the
00:16:32
right
00:16:34
prompt by 2022 AI tools become
00:16:38
mainstream very quickly everyone from
00:16:42
artists to amateurs rush to cash in no
00:16:45
art training needed
00:16:48
imagine making money with art without
00:16:50
actually having to be an
00:16:53
artist
00:17:00
for creating a children's book in just
00:17:03
one weekend using Chat
00:17:05
GPT to using MidJourney to illustrate an
00:17:09
entire graphic novel many artists fear
00:17:13
that there's more to come
00:17:16
regardless of how many decades you spend
00:17:18
honing your skills as an artist
00:17:21
overnight you could become a prompt and
00:17:23
all of a sudden you no longer own your
00:17:26
style if you Google many artist names
00:17:28
you have to be very careful because many
00:17:30
times the results that you get back are
00:17:32
not actual paintings by that artist
00:17:34
they're mimics of that artist style you
00:17:36
know submitted and uploaded by random AI
00:17:39
user somewhere and so this is incredibly
00:17:41
distressing
00:17:42
i think it even took the AI companies by
00:17:45
surprise just how quickly this has all
00:17:48
taken off and gone mainstream graphic
00:17:50
designers are wondering if they're going
00:17:51
to have jobs because if you are just
00:17:54
looking for a decent illustration AI
00:17:56
will do the trick now so these consumer
00:17:58
tools have completely upended a number
00:18:01
of industries already
00:18:13
i'm having such a hard time selling my
00:18:16
work in a competitive market now with so
00:18:21
much fake art and most people don't know
00:18:24
the difference artists and creators are
00:18:27
being harmed by their own work and and
00:18:29
that's just awful on in so many
00:18:33
different ways
00:18:35
the fear within the arts community is
00:18:39
twofold the creative and cultural
00:18:42
industries are afraid that if we don't
00:18:44
adopt generative AI fast enough we're
00:18:47
going to miss the boat we're going to
00:18:48
lose a lot of money the arts and
00:18:50
cultural sector which is visual artists
00:18:52
individual artists content creators the
00:18:55
fear of these people in that sector are
00:18:58
numerous loss of income loss of the real
00:19:03
true meaning of what art might be uh the
00:19:07
loss of
00:19:24
freedom by 2023 the age of AI has
00:19:28
arrived and everyone is hooked in just
00:19:31
over a year more than 15 billion images
00:19:34
are created on AI platforms or over 34
00:19:38
million images a day a number that
00:19:41
continues to grow
00:19:44
for these artificial intelligence
00:19:46
platforms the secret of their success
00:19:48
lies in a database that contains a list
00:19:51
of over 16,000 artists and links to
00:19:54
their creative works
00:19:57
i discovered that these lists that were
00:20:00
being shared in midjourney there's a lot
00:20:03
of long deadad artists but I also see
00:20:06
the names of my peers people who are
00:20:09
still alive and the members in the
00:20:11
Discord encouraging others to try out
00:20:14
combining artists or creating you know
00:20:17
images in the style of artists in the
00:20:21
spreadsheets but how and where did all
00:20:24
these artist names come
00:20:26
from behind this phenomenon lies a data
00:20:29
mining operation that never stops
00:20:32
picture a relentless army of spiders
00:20:35
known as web scrapers their one job to
00:20:38
crawl and hunt the web for data
00:20:42
what scrapers are doing is they find all
00:20:44
of the links to images they turn those
00:20:47
into giant spreadsheets and they say
00:20:49
"Ignore everything about this website
00:20:51
except for the images and just download
00:20:53
as many as they can as quickly as
00:20:55
possible."
00:20:58
Scraping is this simple idea of writing
00:21:00
a piece of software that will go out
00:21:02
online pretend to be a human being and
00:21:04
then basically consume and download data
00:21:07
it's long been used as a research tool
00:21:09
in the late '9s when the internet and
00:21:12
the web was just becoming viable
00:21:14
commercially
00:21:17
in the 2000s pioneering organizations
00:21:21
like Common Crawl began scraping the
00:21:23
internet to archive website pages for
00:21:26
research
00:21:27
purposes people were writing scrapers
00:21:30
just to you know make backups of content
00:21:32
uh the Wayback Machine the Internet
00:21:34
Archive to record the evolution of the
00:21:37
internet
00:21:41
working relentlessly over the years
00:21:43
scrapers have amassed billions of images
00:21:46
audio and texts from users without their
00:21:49
knowledge creating a gold mine of data
00:21:52
for AI companies to train their
00:21:55
models with no laws in place to stop
00:21:58
data scraping anything published online
00:22:01
is fair game over the years it's really
00:22:05
become clear that scrapers are very very
00:22:08
difficult to stop when they're being
00:22:10
used without consent and the
00:22:12
justification is nothing should really
00:22:14
stand you know in the way of
00:22:15
technological innovation
00:22:20
in 2022 Lion the German largescale
00:22:24
artificial intelligence open network
00:22:26
releases the world's largest collection
00:22:28
of image data for public use so Lion 5B
00:22:33
is a data set that contains billions of
00:22:36
links to images and it also has the
00:22:39
captions the alt text for those images
00:22:42
and those together the description of
00:22:44
the image and the image itself are what
00:22:47
powers these machine learning models
00:22:50
although supposedly created for research
00:22:52
and educational purposes the Lion 5B
00:22:55
data set is quickly exploited by
00:22:58
companies to train commercial AI systems
00:23:02
a lot of the AI companies really heavily
00:23:05
lean into the term publicly available to
00:23:09
justify how they got their training data
00:23:13
but anyone who produces content would
00:23:16
tell you just because you see an image
00:23:18
online doesn't give you license to use
00:23:20
the image the Lion 5B database enabled
00:23:25
all of these companies to take and
00:23:29
profit and not even go through legal
00:23:32
means to do
00:23:34
it we have questions like why didn't why
00:23:37
did you speak to a single one of us
00:23:41
meanwhile other concerns have arisen
00:23:43
with web scrapers
00:23:45
before anybody knew it every single
00:23:47
image on the internet was already being
00:23:49
trained into these models so all sorts
00:23:50
of questionable things you know uh
00:23:53
images of of violence illegal things
00:23:56
abuse of all different types incredibly
00:23:58
private images from hospital beds and
00:24:00
and people's private photos all that was
00:24:03
sucked in as long as it was public on
00:24:05
the internet reachable by some crawler
00:24:08
somewhere in December 2023 a controversy
00:24:12
ensues when images of child sex abuse
00:24:14
are found in the Lion 5B database this
00:24:18
forces Lion to remove and review their
00:24:21
data set for safety before releasing it
00:24:23
again
00:24:24
online but some wonder if more could
00:24:27
have been done right from the start
00:24:30
we're going back to a discussion where
00:24:33
AI ethics was not even actually a
00:24:36
discussion and I remember going to calls
00:24:38
from the government where they would say
00:24:40
"What do you mean ethical implications
00:24:41
there are no ethical implications this
00:24:43
is science it works." It's moving so
00:24:46
fast that a lot of people are very
00:24:48
concerned that they haven't had a chance
00:24:51
to even examine what these companies are
00:24:53
doing with their work whether or not
00:24:55
that ends up being legal it already
00:24:57
happened we can't go back in time
00:25:01
in 2023 Hollywood fights back against AI
00:25:05
screenwriters and actors go on strike
00:25:08
raising concerns about the use of AI
00:25:10
tools to write movie scripts and replace
00:25:13
their
00:25:14
likeness some also fear that companies
00:25:17
intend to use AI to reproduce their
00:25:19
faces bodies and voices on screen
00:25:24
i won't say the name of the production
00:25:25
but it was a requirement under the last
00:25:27
contract that I be scanned for that
00:25:29
production so I don't know the uses of
00:25:31
that i did give my permission but I
00:25:33
don't know what the further future uses
00:25:35
of that material my scanning of my
00:25:37
likeness and my voice would be
00:25:41
more alarms are raised in 2024 when
00:25:44
several engineers from Open AI and
00:25:46
Google sign an open letter warning about
00:25:49
the serious risks posed by artificial
00:25:52
intelligence they call for stronger
00:25:54
oversight and transparency
00:25:57
ai companies are
00:25:59
increasingly secretive about what
00:26:01
they've trained on so I think
00:26:03
realistically if if we want AI companies
00:26:07
to tell us what they're training on we
00:26:09
we need to pressure the government to
00:26:12
pressure them to do so
00:26:16
this is a technology that's clearly very
00:26:18
powerful and that we cannot say with
00:26:20
certainty exactly what's going to happen
00:26:21
but it's easy to imagine that it's going
00:26:23
to have like massive effects on the
00:26:25
world and that it could go very
00:26:28
wrong meanwhile some working in the AI
00:26:31
community have begun to take action in
00:26:34
2022 Spawning a startup led by artists
00:26:38
and technologists launches a free
00:26:40
website to help artists find out if
00:26:42
their work is at risk from AI scraping
00:26:46
so have I been trained as a website
00:26:48
where creatives can search the largest
00:26:51
AI training data set for images they can
00:26:55
search for themselves find their work
00:26:56
and then register an opt out which is a
00:26:59
machine readable opt out which we then
00:27:02
communicate to the AI companies to say
00:27:04
"This person does not want you to use
00:27:05
this image in their training." An artist
00:27:09
like that might be Kelly
00:27:11
McCernon we can see that quite a bit of
00:27:14
her work was captured in this Lion 5B
00:27:16
data set we could also look for
00:27:20
Jinga
00:27:23
Jang and uh sure enough there is quite a
00:27:27
bit of her photography here over a 100
00:27:30
images that has appeared in the most
00:27:33
popular data set for AI training and if
00:27:36
she'd like to opt those out she could
00:27:39
then click select all and ask to mark
00:27:42
all of those items as do not train when
00:27:45
she does that AI organizations can check
00:27:48
if any of those uh images that they're
00:27:50
about to download have opted out then we
00:27:52
will tell them that those are off limits
00:27:55
so far have I Been Trained has helped
00:27:57
artists to opt out over 2 billion images
00:28:00
from AI data sets but it has
00:28:03
limitations there are still many other
00:28:06
databases that are likely being used so
00:28:09
the process of opting out is complex
00:28:12
there really is no practical way yet to
00:28:15
remove images from models that are
00:28:18
already trained so really the only thing
00:28:20
that artists can do with the opt- out
00:28:22
mechanism is get themselves out of
00:28:24
future models
00:28:27
meanwhile the ongoing legal controversy
00:28:29
surrounding her copied photo has taken a
00:28:32
toll on Jinga
00:28:34
i have this ongoing copyright lawsuit in
00:28:37
Luxembourg where it's just been eating
00:28:40
away at my life you know where I was
00:28:43
harassed uh doxed with my home address
00:28:46
leaked and even though I was doing
00:28:50
really well in my career before this
00:28:52
happened I had just photographed
00:28:53
Michelle Yo for the cover of the
00:28:55
Hollywood Reporter and I I had no choice
00:29:00
but to pass on all of this opportunities
00:29:04
and give up on my career because I I had
00:29:07
no bandwidth and no mental ability to to
00:29:09
deal with that with her career largely
00:29:12
on hold a major announcement from two
00:29:15
popular art sharing websites soon gives
00:29:18
her a new purpose
00:29:20
when the news broke of platforms like
00:29:24
ArtStation and Deviant Art uh saying
00:29:27
that you know they were going to opt in
00:29:30
uh their users to have their work
00:29:31
scraped for AI training automatically
00:29:35
i was horrified right these are art
00:29:37
platforms and they should try as much as
00:29:40
they could to be protective of their
00:29:42
their artists their users because I saw
00:29:44
that nobody was thinking about this I
00:29:47
ultimately decided that I should step up
00:29:50
and create
00:29:55
KA after weeks of work Jingna and a team
00:29:58
of volunteers launched the beta version
00:30:00
of KA in 2022 it's a social media
00:30:04
platform where artists can showcase
00:30:06
their
00:30:07
creations unlike other sites their
00:30:09
artwork will not be opted in for AI use
00:30:14
so currently there is no technology on
00:30:16
earth that could prevent someone's work
00:30:19
from being scraped uh by AI like if you
00:30:22
post it online it could be scraped right
00:30:24
so what we try to do is the most basic
00:30:27
thing that all social media and art
00:30:30
platforms should be doing which is to
00:30:32
opt people out of being scraped by AI by
00:30:35
default because opting people in by
00:30:38
default is
00:30:40
wrong car's mission to host an AI free
00:30:43
art community is met with a mix of
00:30:45
curiosity and caution but Jinger is
00:30:50
undaunted for everyone else from tech
00:30:53
giants to big brands though the race for
00:30:56
more data has already
00:30:58
[Music]
00:31:01
begun from films to music cataloges
00:31:05
creative assets are now being openly
00:31:07
sold to AI companies who are eager for
00:31:09
more content to train their machines on
00:31:14
[Music]
00:31:19
and it remains to be seen if anyone can
00:31:22
do anything about
00:31:38
[Music]
00:31:42
it in the spring of 2023 an
00:31:45
unprecedented event shakes up the world
00:31:48
of art a portrait named the electrician
00:31:51
wins a Sony World Photography Award one
00:31:53
of the most prestigious prizes in the
00:31:55
field but no one including the judges
00:31:59
knows that this is AI
00:32:03
generated the picture was created by
00:32:06
prompttographer Boris El Dagon people
00:32:09
are attracted to the image because it
00:32:11
has an emotional quality many people
00:32:14
criticize AI for having no emotion
00:32:18
having no soul but that image shows that
00:32:21
it can have it depends on the person
00:32:24
producing the
00:32:26
image the people do with AI is often
00:32:29
very futuristic and science fiction and
00:32:33
this is what I liked about Boris work it
00:32:36
it seems like pictures taken in the
00:32:38
Second World War or in the
00:32:41
1940s and that's why I think he won the
00:32:44
Sony awards because the jewelry did not
00:32:47
realize that it was not a real picture
00:32:50
you know
00:32:52
revealing that he used AI at the awards
00:32:54
ceremony boris refuses to accept the
00:32:57
prize he argues that he submitted the
00:33:00
image for a reason to provoke a
00:33:02
discussion about the use of AI in
00:33:04
photography
00:33:07
my mentality was uh one of a hacker it
00:33:11
was a kind of a prank here because that
00:33:13
was the necessity i saw a chance to do
00:33:16
something good for the photo community
00:33:19
that I'm part of
00:33:22
as more and more people adopt the
00:33:24
technology pundits like Boris feel that
00:33:27
people have no reason to fear AI if
00:33:30
you're not using AI tools in 2025 my
00:33:33
friend you are missing out hi everyone
00:33:35
I'm Ishan Sharma and in this video I'll
00:33:38
share with you seven amazing AI tools
00:33:40
that will help you save time make more
00:33:42
money and have a better quality of life
00:33:45
ishan Sharma is a popular business
00:33:48
YouTuber in India who relies on AI to
00:33:51
help produce
00:33:52
content an AI evangelist he believes the
00:33:55
positives of using it far outweigh its
00:33:59
faults i know what many of you may think
00:34:02
AI art should be legal since it's
00:34:03
stealing artwork of hardworking artists
00:34:06
my take is to understand how AI works we
00:34:08
need to first understand how humans
00:34:09
create things and what is creativity
00:34:12
from scratch creativity in simple terms
00:34:14
is you mixing all of your experiences to
00:34:16
make something new and original we've
00:34:18
been mixing multiple concepts and
00:34:19
beliefs for centuries to make new
00:34:21
reforms arts rules etc and AI is doing
00:34:24
nothing different i think it's the same
00:34:27
situation with Ford so when Ford
00:34:29
introduced its first cars there were
00:34:30
people using horse cars to actually
00:34:33
travel they changed from the horse cars
00:34:35
to the to the actual cars and they were
00:34:37
able to save time and do things faster
00:34:39
to all the people who are watching this
00:34:40
video who are afraid of generative AI in
00:34:42
general I would suggest you to just use
00:34:44
it once and see how it can help you
00:34:47
accelerate in your work in your
00:34:49
day-to-day
00:34:52
life in May 2024 Jingna wins the appeal
00:34:56
in her copyright dispute against
00:34:58
Luxembourg painter Jeff Dishber the
00:35:01
court rules that her work was copied
00:35:03
without her consent
00:35:06
i feel really relieved and glad and
00:35:10
grateful it also means a lot to
00:35:13
photographers and artists that you know
00:35:16
copyright is being upheld and I hope
00:35:20
that this can also aid in regulations
00:35:24
and legislations being passed to protect
00:35:27
creatives especially as AI become more
00:35:30
and more widely adopted across various
00:35:32
industries
00:35:36
while Jinga's victory marks a turning
00:35:38
point the battle is far from over for
00:35:40
the creative
00:35:43
industry at a US Senate hearing over AI
00:35:46
leading advocate Carla Ortiz becomes the
00:35:49
first artist to outline to the
00:35:51
government the potential harms of
00:35:53
technology these
00:35:55
work and used to train for As the debate
00:35:59
over the issue grows she launches a
00:36:02
class action lawsuit with fellow artists
00:36:05
against AI companies like Open AI and
00:36:08
Stable Diffusion for infringing on their
00:36:11
copyrights jingner also decides to join
00:36:14
them i have known Color Autist for many
00:36:17
years she knew about the copyright
00:36:19
lawsuit that I've had going on for the
00:36:21
last 2 years so ultimately I thought it
00:36:25
was really important to be part of this
00:36:27
i want to take the opportunity to show
00:36:29
that no is it's actually not okay
00:36:32
so currently we have 10 artists who are
00:36:36
suing um these generative AI companies
00:36:39
for the use of our work in their
00:36:41
machines i know this is a marathon and
00:36:44
we're going to be doing this for many
00:36:45
years things are about to really go
00:36:47
uphill get a lot tougher but this is
00:36:50
what we want
00:36:52
with the laws still unclear on governing
00:36:55
AI though the road to justice won't be
00:36:58
an easy one a lot of the lawsuits
00:37:00
against the AI companies are arguing
00:37:02
that the data is being illegally copied
00:37:05
the AI companies will tell you that
00:37:07
technically they're they're not storing
00:37:10
a copy so it's not really a reproduction
00:37:13
so I think the way that AI companies are
00:37:16
going to respond to these complaints is
00:37:18
to assert fair use they're going to say
00:37:21
that the models are transformative they
00:37:24
don't output exact copies of the inputs
00:37:27
what these models are learning are just
00:37:28
statistical patterns out of these images
00:37:30
and not the images
00:37:36
themselves trust issues remain as tech
00:37:39
giants continue to scrape user data by
00:37:42
getting consent via their terms of
00:37:44
service each time anyone logs on
00:37:47
in 2024 Meta declares that all user
00:37:51
content posted on Facebook and Instagram
00:37:54
will also be used to train
00:37:57
AI overnight these terms trigger an
00:38:00
exodus as hundreds of thousands of
00:38:02
artists sign up for KA the anti-AI
00:38:05
platform started by
00:38:09
Jingna jingazang creator of the new art
00:38:11
and photosharing social network cara
00:38:14
posted new users on the platform had
00:38:16
crossed a 100,000 a few days later
00:38:19
300,000 new users and then just
00:38:21
yesterday a half a million and artists
00:38:23
are leaving Instagram in droves for this
00:38:25
new
00:38:30
site elsewhere technologists like Ben
00:38:34
Jao are coming up with innovative
00:38:36
solutions designed to protect the works
00:38:38
of artists from data scraping
00:38:42
i run a lab uh co-directed with my
00:38:44
colleague Hedo Jung and you know we have
00:38:47
a lab of a few graduate students PhD
00:38:49
students we spend our days thinking
00:38:51
about how to mitigate harms of
00:38:53
generative AI on human beings
00:38:56
the team's expertise in building cyber
00:38:59
privacy tools would prove to be a
00:39:01
gamecher for artists against
00:39:05
AI they develop a software called Glaze
00:39:09
it adds a layer on the artwork that
00:39:12
confuses web scrapers and prevents AI
00:39:15
from copying details of the original
00:39:18
image think of it like if you can
00:39:20
imagine putting on UV glasses that
00:39:23
enables you to see an image in a
00:39:26
complete different spectrum or frequency
00:39:28
right so a set of images that for
00:39:31
example start as charcoal portraits very
00:39:34
dark and moody and low contrast for
00:39:36
example um can be altered by glaze you
00:39:39
know through a digital transformation
00:39:40
process you feed it in it changes a
00:39:42
bunch of pixels out comes a separate
00:39:44
image a second image will be altered in
00:39:46
such a way that even though it still has
00:39:48
that same almost exact same look but AI
00:39:51
will look at that altered image and see
00:39:53
something completely different instead
00:39:54
of a dark charcoal portrait you know
00:39:58
ideally you wanted to see like cartoony
00:40:00
super vibrant colors
00:40:04
artists can download glaze on their
00:40:06
computers to treat their images before
00:40:08
posting them online more artists learn
00:40:12
about the software through KA the first
00:40:14
platform to integrate Glaze it's a sign
00:40:18
of more solutions to come
00:40:24
[Music]
00:40:29
so today we are at Lightbox Expo in
00:40:32
Pasadena and it's a 3-day event for
00:40:35
industry artists in films video games
00:40:38
animation some illustration to kind of
00:40:41
come together listen to panels and we
00:40:43
have a booth for KA
00:40:45
here people have been coming up to us
00:40:48
and it's been very nice to talk to
00:40:50
people face to face i like that we exist
00:40:52
that that we made an impact on someone's
00:40:54
life that's meaningful you know i just
00:40:57
feel very optimistic seeing other
00:40:59
artists take action because I think it
00:41:01
gives us hope i think artists often they
00:41:04
they can feel very small as individuals
00:41:07
and actions like creating car so I I
00:41:10
think what she did was just incredible
00:41:13
i'd personally like to really thank
00:41:15
Jingna uh Carlo Ortiz Sarah Anderson at
00:41:20
Lightbox Expo this year because they try
00:41:24
to make a difference as this new
00:41:27
technology of generative AI is coming in
00:41:30
without
00:41:31
them there wouldn't be really anybody to
00:41:34
help fight for us
00:41:37
some tech startups are now hoping to
00:41:40
create a more positive relationship
00:41:42
between the creative industry and AI
00:41:44
companies one built on consent and fair
00:41:47
compensation
00:41:49
the source plus is a marketplace so
00:41:53
creatives can go and offer their data
00:41:56
for sale ai companies will be able to go
00:41:59
and get opted in data that compensates
00:42:02
the creators we do think it's possible
00:42:04
for image models to be trained without
00:42:07
infringing on artists
00:42:11
these new initiatives are giving the
00:42:13
creative community hope but even as they
00:42:16
continue to find solutions others feel
00:42:19
that more needs to be
00:42:21
done in a landmark decision in 2023
00:42:25
China becomes the first country to
00:42:27
recognize copyright in AI generated art
00:42:31
[Music]
00:42:33
vote is closed and it is adopted
00:42:39
congratulations then in August 2024 the
00:42:42
EU approves the world's first
00:42:45
comprehensive legal framework to
00:42:47
regulate AI acknowledging the risks of
00:42:50
manipulation and
00:42:52
misuse what is created by AI should be
00:42:56
identifiable so in this way we can give
00:43:00
some support to authors to creatives
00:43:02
that need to know when creative material
00:43:05
that they produced was used to train the
00:43:07
AI so that they can check if the
00:43:09
copyright law was respected or if it was
00:43:14
violated but in the US and other
00:43:16
countries the laws are still evolving
00:43:19
and the debate over who owns AI
00:43:21
generated art continues
00:43:25
an AI model is an imitation of human art
00:43:28
without a human teaching them they have
00:43:30
no idea what is good art so if you take
00:43:32
human art out of the equation then they
00:43:34
are just copying from other copies and
00:43:36
the only question is where is the human
00:43:39
artist going to be in 20 years what's
00:43:41
happening in some of the art schools our
00:43:43
students are dropping out because they
00:43:45
don't want to be affected by AI and
00:43:47
there's going to be very little human
00:43:48
art to help AI learn in the future an AI
00:43:51
is just gonna be stuck learning from its
00:43:53
own copies and just getting worse and
00:43:55
worse and worse
00:43:57
generative AI really represents a threat
00:44:00
to art and to culture not in the sense
00:44:02
of just loss of income but also once it
00:44:07
is adopted more broadly across society
00:44:11
that's when it starts to represent a
00:44:14
real risk in terms of what stories we
00:44:18
read and what images that surround us
00:44:21
and how we think because it can become a
00:44:24
way of standardizing how we express
00:44:28
ourselves and even manipulating public
00:44:31
opinion by a handful of companies
00:44:35
for artists and writers the loss of the
00:44:38
human touch with the ongoing adoption of
00:44:41
generative AI is perhaps the greatest
00:44:43
concern of all
00:44:46
the AI companies and the AI researchers
00:44:48
don't think of themselves as you know
00:44:50
villains sneaking in and and stealing
00:44:53
data they think of themselves as uh
00:44:56
researchers who are sharing all of the
00:44:58
tools that that are creating these quite
00:45:00
magical systems these folks actually do
00:45:03
think of what they're doing as as
00:45:04
impactful for the future of humanity in
00:45:06
some cases they're right what they don't
00:45:09
understand though is that we have
00:45:10
artists who uh are going to very likely
00:45:14
lose work because of these systems and
00:45:16
if we were able to better communicate I
00:45:18
think we would have a lot better uh
00:45:20
solutions to these problems
00:45:28
[Music]
00:45:36
purchase
00:45:38
[Music]
00:45:59
[Music]
00:46:27
[Music]
