AI vs Artists - The Biggest Art Heist in History

Summary

The video discusses the impact of AI image generators like Midjourney on artists. It explores concerns around copyright infringement and data scraping, with models trained on billions of images without consent. The speaker argues this is hurting artists directly by allowing easy mimicry of unique styles. They examine ethical issues like impersonation and generated fake content. Overall, they advocate careful, selective use of AI art to avoid jeopardizing creative futures or discouraging emerging talent through over-automation. The video calls for transparency, consent, and compensation around training data as well as human review of AI art laws.

Takeaways

• 😮 Generative AI models like Midjourney are trained on billions of images scraped from the internet without consent, including artists' copyrighted work
• 😢 Artists feel their unique styles are being stolen by AI to generate images, while they receive no compensation or credit
• 😠 There are no regulations restricting companies from scraping copyrighted content for AI training data
• 😕 AI-generated art saturates online spaces, making it harder to trust content and appreciate human creatives
• 😊 Ethical models like Adobe's train on stock image libraries with creator consent and compensation
• 👀 Tools like Nightshade and Glaze can help creators poison-pill images to protect against scraping
• 🤔 Contributing images to AI models may enable style mimicry and involuntary data laundering
• 😔 Young artists feel hopeless seeing AI instantly achieve aesthetics that took them years to develop
• ⚖️ Clear legislation is needed around transparency, consent, and compensation for AI training data
• 😌 Human creativity, skill, and vision will likely regain dominance as AI art becomes industrialized junk food content

Q & A

What are some of the main concerns artists have regarding AI image generators?

Artists are concerned that their work is being scraped from the internet without consent and used to train AI models. This allows companies to profit from their work without compensation or credit. There are also concerns about potential reputation damage, forgery, and identity theft.

How does AI art affect young and aspiring artists?

Many young artists feel hopeless and discouraged because AI can generate quality artwork with little effort, making them question spending years honing their skills. This could negatively impact the number of people who pursue art as a career.

What are some solutions proposed to make AI art more ethical?

Solutions include only using work from artists who consented, providing royalties and compensation, cataloging generated images to make them traceable, banning direct style mimicry, and requiring a mix of at least two art styles.

How does AI art affect perception of what is considered quality art?

Seeing a lot of AI art online will make people assume most art is AI-generated. This means great human artists may not get recognition, as people doubt the authenticity of art from new and emerging artists.

Why can AI art not be copyrighted in its current form?

The last step of creative expression is done by the AI model based on the prompt, not the person typing the prompt. Without human authorship and creativity, AI art cannot be copyrighted under current laws.

What are some flaws in believing AI art democratizes art?

Most AI artists want to copyright their own work and profit from it commercially. The art community has always welcomed beginners, so there have not been major barriers to entry other than dedication and practice.

How does generative AI threaten the appreciation of art?

AI art is fast, cheap consumerism rather than something contemplated and appreciated. Losing the connection between art and the artist's skills, efforts, experiences and feelings removes meaning and value.

What are examples of companies taking an ethical stance on AI art?

Some companies like Procreate explicitly avoid integrative AI to focus on empowering human creativity. Others like Adobe and Shutterstock are transparent about using only consented data in training.

What skills do human artists have that AI lacks?

Humans have skills AI lacks like true creativity, surprise, conveying thoughts and feelings, and the ability to make glorious mistakes that lead to breakthroughs. This will keep human art valuable.

How can artists with influence make a difference regarding AI art?

Influential artists can speak out about concerns over losing the option to pursue art as a career, and discouragement of young artists. They can value human skill and creativity in their own work.

Outlines

00:00

🎨 Generative AI's Impact on Art and Copyright

This section highlights the concerns of artists and creators about generative AI, such as MidJourney and OpenAI's Sora, which create art and videos from text prompts. It discusses the ethical and legal issues surrounding the use of massive datasets like LAION-5B, which contains billions of images without the consent of the original creators, to train AI models. The potential for these technologies to undermine copyright laws and the creative process is a significant concern, as AI can replicate styles and produce art that competes with human artists, raising questions about originality, compensation, and the future of creativity.

05:01

🔍 The Business Perspective and Ethical Concerns in AI

This paragraph addresses the predominantly business-first approach in discussions about AI, highlighting the need for a more inclusive perspective that includes artists and creators in the decision-making process. It critiques the practices of companies like Mid Journey for using copyrighted images without permission, likening the impact of generative AI on artists to the disruptions caused by file-sharing services on the music industry. The emphasis is on the potential of generative AI to not only use artists' work without compensation but to replace them entirely, raising ethical concerns about the sustainability of creative professions.

10:02

👥 The Personal Toll on Artists and Style Mimicry

Focusing on the personal stories of artists affected by AI's ability to mimic styles, this paragraph illustrates the emotional and financial impact on creators who see their unique styles replicated without consent or compensation. It introduces the concept of 'consistent styles' by platforms like Mid Journey, which exacerbates the issue by allowing users to generate art in any artist's style, further blurring the lines between original work and AI-generated content. The discussion also covers the inadequate responses by AI companies to copyright infringement, highlighting the broader implications for artists' rights and recognition.

15:03

🤖 Ethical Dilemmas and the Future of Artistic Creation

This section delves into the ethical quandaries posed by generative AI, including the manipulation of data to create art that mimics specific styles, raising questions about originality and copyright. It discusses the impact of AI on the value of human creativity and the potential for AI to saturate the market with content that lacks the depth and authenticity of human-made art. The text also examines the implications for emerging artists and the importance of ethical considerations in the development and use of AI technologies in creative fields.

20:04

🌐 The Proliferation of AI-Generated Content and Copyright Issues

Exploring the widespread distribution of AI-generated images and their impact on digital platforms and stock image libraries, this section highlights the challenge of distinguishing between human and AI-created content. It points out the transparency efforts by companies like Adobe with its Firefly model, trained on a unique dataset, and contrasts this with the broader issue of non-consensual use of copyrighted materials. The discussion extends to the implications for copyright law, artist compensation, and the need for ethical standards in AI-generated content.

25:06

👓 Perception of Reality and the Value of Human-Created Art

This paragraph reflects on how generative AI affects the public's perception of art and reality, suggesting that the abundance of AI-generated images could numb people's appreciation for the intricacies of human-made art. It emphasizes the importance of the creative process and the unique experiences and emotions that artists bring to their work, which AI cannot replicate. The text argues that while AI art may offer accessibility and instant gratification, it lacks the depth and personal touch that characterize human creativity, potentially leading to a devaluation of genuine artistic talent and effort.

30:09

🚀 The Debate on Generative AI's Ethical Use and Copyright Reform

Discussing the ongoing debate over the ethical use of generative AI and the need for copyright reform, this section outlines the views of various stakeholders on how to address the challenges posed by AI in creative industries. It explores potential solutions like consent-based training datasets, compensation for artists, and stricter copyright laws to prevent data laundering and style mimicry. The text underscores the importance of transparency, accountability, and responsible innovation in the development of generative AI tools to ensure they complement rather than compromise human creativity.

35:11

🌟 Prospects for Ethical Generative AI and Creative Futures

Looking towards the future, this section examines the potential for developing ethical generative AI that respects artists' rights and contributes positively to creative industries. It discusses the efforts of companies like Adobe and initiatives like the Content Authenticity Initiative to integrate ethical principles into AI development. The paragraph emphasizes the importance of legislative action, community standards, and technological solutions to ensure that AI serves as a tool for enhancing human creativity rather than undermining it, preserving the opportunity for future generations to pursue art as a viable career.

40:13

🎭 The Importance of Human Creativity in the Age of AI

Concluding the discussion, this paragraph champions the irreplaceable value of human creativity and the unique qualities that human artists bring to their work, which AI cannot replicate. It reflects on the role of discipline, passion, and hard work in the creative process, using the example of PewDiePie's art journey to inspire and encourage artists. The text argues against the notion that AI can or should replace human creativity, advocating instead for a future where technology supports and enhances the artistic expression, ensuring that art remains a meaningful and vibrant part of human culture.

Keywords

💡generative AI

Generative AI refers to artificial intelligence systems that can generate new content, such as images, videos, text, etc. The video focuses extensively on image generation systems like DALL-E and Midjourney that can create realistic images based on text prompts. It discusses concerns around these systems being trained on copyrighted or inappropriate data as well as legal and ethical issues around mimicry of artist styles.

💡training data

The data used to train generative AI models - often large datasets of images, text, videos etc scraped from the internet without much curation. The video argues this training data likely contains copyrighted art, inappropriate content etc and that artists should have consented and been compensated.

💡data laundering

The process of taking data that was scraped without permission, transforming it (like compression) and then claiming it is legitimate to use to train AI models commercially. The video argues this is how stolen artist work ends up training generative models.

💡artistic style

The distinctive visual style of a particular artist. The video discusses AI concerns around models not just copying styles but also perfectly mimic specific artists using few reference images, potentially damaging their careers.

💡consent

Permission given by artists for their work to be used in AI training data and models. The video argues consent is not properly taken from artists currently.

💡copyright

The legal ownership and rights over creative work. Copyright law does not properly cover AI-generated mimicry of artist work currently. The video recommends updated legislation to empower artists.

💡compensation

Payment to artists for commercial use of their work in training AI models and generating profit from their styles. The video argues proper compensation currently does not occur.

💡transparency

Being open about what data is used to train AI models. The video recommends model training datasets should be fully transparent about included data.

💡ethics

Principles of morally right behavior. As AI artwork increases, the video discusses dilemmas around ethical use of these models and data vs profit-seeking at any cost.

💡regulation

Laws and government oversight governing use of new technology like AI. The video argues better regulation is needed around generative models and data practices.

Highlights

AI models are designed to generate new images similar to what they have seen during their training.

The more distinct an artist's style is the more successfully AI can mimic it.

The real appreciation of art comes from valuing the process of the artist, not just the final work.

Looking at generated images is numbing your senses and distorting your perception of reality.

Relying mainly on generative AI to make money is not going to last long even if the creator tries to hide the AI use.

Without an AI tool, most AI artists wouldn't be able to even come close to their generated images.

The companies behind social platforms are baiting users with pretty AI pictures and videos, harvesting their attention and selling it to advertisers.

Generating images, videos, music, 3D objects and everything else will quickly lose its novelty and hype.

The real danger of generative AI is its effect on up-and-coming artists. Many talents may have already been lost to discouragement.

I want a future where there's an option to pursue art as a career. I want kids to want to draw, want to create, want to make things that mean something to them.

Drawing is a visual language, a form of communication. Just because a robot can speak doesn't mean we should stop speaking.

Generative AI models are devouring art in order to mimic greatness created by artists, but they will never have a soul or feelings to express.

At some point human art is going to be valued higher than AI art. AI will always be second rate.

There will be a constant race to identify real human creativity versus something that just came out of a pixel predictor.

We're putting the whole creative future of humanity at risk with generative AI because it's not just skipping fundamentals, it's skipping the emotional journey of learning.

Transcripts

I've seen a lot of my friends just

suffering because their work is so

prominent and they struggled for 10

years or more trying to get their Style

just to see someone else parading

wearing their face and parading it right

there is nothing stopping ad companies

from taking whatever you own from the

internet downloading it and putting into

their training model nothing really

we're putting the whole creative future

of humanity at risk here in the last

couple of weeks I spoke to many amazing

artists and scientists about my mixed

feelings of generative AI join me to

hear their thoughts my advice to

creators and predictions on what's to

come I will be perfectly honest with you

I am equally fascinated and terrified by

generative AI as a graphic designer

illustrator and educator I try to stay

on top of any new technology and tool

that helps self-expression and

creativity I still remember the first

time I saw the potential in AI use for

creating art it was in 2018 when I saw

the example Scott Eaton achieved using

his model trained exclusively on his own

photography data set from his body's in

motion project I never thought that by

2024 we will have tools like mid Journey

that can generate photorealistic images

indistinguishable from real photos or

art and tools like open AI Sora capable

of generating ultra realistic videos

from Tex prom

by now most likely you are familiar with

how AI images are generated but in case

you are not let me give you a very brief

and extremely simplified summary you

type in a few words your prompt

describing what you want to see and the

AI model generates images for you that

never existed before for this process to

work well the AI module needs to be

trained on a massive amount of images

and their descriptions at the moment the

most effective and most widely used data

set Layon 5B includes

5.85 billion uncurated images cwed from

publicly available websites and Cloud

storages this data set was meant for

research purposes only but companies

started to utilize it commercially

within a few months after its release

there are all kinds of images in this

data set from public domain pictures to

a staggering amount of copyrighted work

of both dad and living artists and a

fair amount of explicit content too the

creators of this data set didn't get

consent from anyone while collecting

these images there was no opt-in or opt

out but since it was meant for research

purposes only it didn't seem to violate

anyone's rights you can get a better

idea of the effectiveness and sheer

amount of scraping that was used to

generate Layon 5B data set by going to

have iben tred.com website you can

search the entire training data using

someone's name a domain name or even a

keyword for instance you can find out

how many images of elephants were

scraped off the internet and you can

start to see why AI models can generate

images of elephants from any angle and

in any artistic style this is like

scanning someone's brain to see how well

they remember how an elephant looks only

artists with extraordinary visual memory

like the late Kim jangi can come close

to this level of familiarity and

comprehension of a subject matter

without going into the technicalities of

how deep learning and stable diffusion

works here are some important things to

understand about text to image

generative AI models they are designed

to generate new images similar to what

they have seen during their training

once a model is trained on an image it

cannot unsee it it is a bit like human

memory but much more accurate and

extensive a bit like photographic memory

paired with unlimited storage capacity

to improve the speed of the generation

process all images that are fed into the

AI model are compressed now this

compressed form of the image or latent

image looks like noise to us but the

model is capable of reversing it back to

a recognizable almost identical image as

the original input this transformation

between the original image and the

recovered image after the def Fusion

process is crucial as this is used as an

excuse by AI companies to get around

existing copyright laws and to allow

them to cheat artists out of their

rightful

compensation you may have heard of the

term data laundering it involves

transforming stolen data so that it can

be used for legitimate purposes in this

instance the original data 5.85 billion

images was collected from the web by

nonprofit research groups and then

shared with for-profit companies most

people who are having the AI discussion

are unfortunately starting from a

business first mindset so all of the

ideas all of the discussion about what

should be legal what should not be legal

our discussions are sort of framed

around companies that want to scale up

to the millions to billions of dollars

of Revenue right and if if you're going

to be that kind of company I don't think

you can just have a charismatic CEO at

the top who thinks that they have

perfectly understood the future of

technology and that for some reason

thinks their fairh handed handling of

this situation is going to assure the

best outcomes for everybody no you need

boards that are made up of actual

artists right not just um not not just

technology people who stumbled upon a

code that happens to make art you need

people who are really sort of outside of

the situation real stakeholders um to

help guide these organizations creators

of mid Journey one of the most popular

text to image generators have been

recently CAU discussing laundering

copyrighted images whilst training their

AI model many scientists who worked on

developing AI Technologies are also

disillusioned by the predatory practices

of AI companies used to harm individuals

and entire Industries there is nothing

stopping a companies from taking

whatever you own from the internet

downloading it and putting into their

training model nothing really uh there's

no legal uh statue standing in the way

there's no regulatory agencies that that

say this is not allowed but why is it

such a big deal you might ask that AI

modules have been trained on copyrighted

materials how does this affect the

original

artists remember what happened to the

music industry when nepsta first came

out and then with pretty much every

other intellectual property once fire

sharing or torrent sites became

available I'm talking about movies TV

shows books software Games Etc although

pirated content is still widely

available most Industries found ways to

combat it mainly by streaming their

content the main difference however is

that while torrent sites have been

hurting mainly large companies and

Brands generative AI is hurting

individuals what is worse is that

generative AI didn't just steal artists

work it was created in order to replace

them since we are comparing AI systems

to torrent sites it's as if you can only

download the last season of your

favorite show but even generate any

number of new episodes or Seasons

without the consent and input of the

shows creators if they if they were

marketing a commercial product that was

solely based on people who opted in they

wouldn't have a very competitive product

they wouldn't have a product that would

compete directly with artists right and

like you know this they want to tell

people it's a tool and it's a tool for

artists but it's not it's it's meant as

a replacement for artists because

they're trying to directly compete with

artists in the same market right that is

not a tool that is a replacement the

only way to have a very competitive

extremely competitive product that has

the best imagery is to scrape from

everyone and anyone one of the main

arguments of people using generative AI

tools is that their generated images

only look similar to existing real

artwork that was used to train the AI

models they compared this ability of

mimicking other people's work to how

artists find inspiration and references

here is what l v bar or Lo wrote about

this many have compared image generators

to human artists seeking out inspiration

these two are not the same my art is

literally being fed into these

generators through the data sets and

spat back out of a program that has no

inherent sense of what is respectful to

artists as long as my art is literally

integrated into the system Ed to create

the images it is commercial use of my

art without my consent there are

dedicated websites like this one to help

people easily copy any artist style

using mid Journey you can consider these

style cataloges for people who are lazy

to even research and get to know the

artist they would want to copy Bobby CH

put it this way in his video ai's

ability to copy my style is kind of like

after I spent all this effort to climb a

mountain once I get to a top everyone

else just pops up with the you know

taking selfies on top of this mountain

right beside you and you know that if

you didn't climb the mountain then none

of the others could either yet the only

compensation that's given is to the AI

company even though without my artwork

as well none of this would be possible

they're just looking at something and

they're they're thinking that looks cool

um who did that I'm going to keep

prompting stuff by lowish right then

that's not you're not learning anything

you're just kind of you know it's cool

you don't understand why it's cool and

you just kind of want to keep generating

more of it so that that to me feels like

you're not really benefiting from

anything you're kind of just you know

pulling the slot machine for more for

more results but I've seen a lot of my

friends just suffering because their

work is so PR and they struggled for 10

years or more trying to get their Style

just to see someone else parading

wearing their face and parading it right

you might wonder what happens if you

want to mimic a specific less known

artist style and M Journey doesn't

recognize their name when you're using

it in your prompts well mid Journey

recently introduced a feature called

consistent Styles which makes it even

easier to mimic anyone's style let me

demonstrate this quickly here are some

illustrations my good friend George ton

created

using one or two of these as style

references mid Journey can closely

emulate George's style and the whole

process takes less than a minute within

an hour anyone can generate hundreds of

detailed illustrations in George's

unique artistic style what this means is

that even if his work wasn't scraped in

the first place and used to train the AI

model the uploaded style references will

help to identify similar artwork in the

database and use that to generate

something very similar I don't know if

the uploaded images by the user are then

automatically added to the AI models

training data but I'm pretty sure they

are used in some way to improve it if

that is the case then users become

complicit in data laundering without

their knowledge here is what Patrick

Brown has to say about style mimicry

when I confronted him with generated

images using mid Journey's consistent

Style featur

when I saw this I was kind of surprised

very shocked each of these were probably

just made within seconds compared to if

I was to draw something like this maybe

like this one I would spend probably at

least 3 days probably pulling this

together maybe four or five including

the background for sure like that

background very detailed you know it

does look very

aiish but to the average eye I think it

would just pass as a the usual kind of

Spider-Man artwork or just a piece you

would have thought that it would have

been done by an artist so it's very

scary and I don't like that side of AI

to be honest especially when you aim at

a certain artist and say Hey I want to

get this style and then create a full

picture and out of thin air it really

does suck to have your name or your art

style kind of siphoned off and like put

through a filter and generated and spit

it out in few a few seconds you know

there is a short paragraph about

stealing someone's intellectual property

in mid Journey's terms of service if you

knowingly infringe someone else's

intellectual property and that cost us

money we are going to come find you and

collect that money from you we might

also do other stuff like try to get a

court to make you pay our legal fees

don't do it Sam does Arts explained the

dangers of AI imitating anyone's work

when AI Generations are made to look

like an artist's work this has the

potential for reputation damage for

forgery for fraud for identity theft and

what's most concerning is that when

these models are trained on images of

your artwork they are unable to forget

so almost all of these models are now

working with tainted data their

Generations now all involve copyrighted

content which has been gathered without

the knowledge or permission of the

copyright owner it is directly hurting

artists who have put their passion and

their soul into everything that they

create only for their work to be scraped

from the internet without their

permission and used in training AI

models the unfortunate Paradox is that

the more distinct an artist style is the

more successfully AI can mimic it

Additionally the more complex and

detailed an artwork is the more likely

people will think it is made by AI even

when it's not so generated images

mimicking someone's style is not only

plagiarizing the artist but also makes

people doubt the authenticity of any

future or past work created by the

artist I've been looking at at for I

don't know the past 20 years or so in my

life and yeah sure sometimes there are

dead giveaways about like oh it's too

smooth or does this and that but like

for what you're showing here especially

with their consistency update it's it's

scary like it's it's truly scary you

take someone like gabs for example and

and you just want to emulate his style

you want it for a cover you want it

for an illustration for a movie poster

right you just you take a few more

examples double it triple the uh the

amount that you that you use for

consistency and all of a sudden you have

even better results so I think it's

tremendous tremendously scary because

it's so good this AI ethics Spectrum

demonstrates really well how exact style

mimicry is ethically and morally the

worst thing you can use generative AI

for especially when used to profit

commercially from the generated images

Dana Executive Vice President general

counsel and chief trust officer at Adobe

said this in his interview with neai

Patel from the word there's no such

thing as style protection in copyright

law right now that you can't protect

that there's that's not a concept so

we've uh We've introduced in the same

sen of testimony you refer to earlier

this idea of a federal anti-

impersonation right and the idea of that

is it would give artists a right to

enforce against people who are

intentionally impersonating their work

for commercial gain is it still possible

to share anything online without

worrying about feeding a generative AI

model with your work Ben Xiao professor

of computer science and his team at the

University of Chicago created two

interesting tools that can help protect

your digital images stored online so n

shade is trying to help give some power

back to the to the content owners um

against people who don't care about

licensing who don't care about ownership

who don't care about copyright uh it is

a small poison pill you can put inside

your art um it works very similar to

glaze except where glaze is modifying

what the AI model sees in terms of

artistic style Nightshade works by

changing what the AI model sees in terms

of composition so instead of you know

seeing a cow in a you know Green pasture

it sees a big leather handbag uh sitting

on a grassy Hill right so um each art

piece can be equipped with a small

poison pill um in such a way that if

enough of it gets collected and gets

trained by the same model then that

model will start to get very confused

about what is a cow it will think that a

cow has a nice leathery handle and you

know shiny buckles um and those kind of

things AI is not only saturating the

internet with unreliable fake content

that's free to access but also in the

form of paid content here is an example

of a fake art book that is advertised as

a compelling and extensive guide book on

drawing people while in reality it is a

collection of generated images and

generated text with close to zero value

to anyone interested in learning to draw

fellow YouTuber Jaz dros made a video

about this it just goes to show the

moral ethic iCal ambiguity of AI now uh

and how it's how it's impacting artists

I could have paid that $95

us to the people who whose style that

was based on or or it was used you know

I might have bought J Scott Campbell how

to draw book you know which I and I've

bought some of those I am pretty sure

this is just one of many examples of fig

books out there just imagine you are

learning to drive and you get scammed to

learn from a fake book or you get a

story book for your child that turns out

to be AI generated with hidden flows

that will secretly influence your

child's imagination or understanding of

the world oh wait this already happened

how can you protect yourself from fake

books only by being extremely Vigilant

but it is going to get harder and harder

to differentiate generated books from

real ones as AI models and workflows

improve but it's not just books there

are all kinds of generated digital

products already out there sites like

art station are full of stuff like this

massive collections of image references

of something very specific like a

stylized werewolf head there is a small

mention that this product is made with

the help of AI meaning it's nothing more

than a library of curated generated

images at least here they were not lying

about the use of AI stock sites are also

Ed with generated images to see the

scale of proliferation it's worth

checking the ratio between real images

compared to generated images on some of

the popular stock sites shut stock for

instance currently has around 5 million

AI generated images which sounds like a

lot but it only amounts to

1.15% of its entire Library compared to

this free pick went all in with AI

images a staggering

42.7% out of their 121 million stock

images are generated I have to give it

to them though that they are being

completely transparent about this and

they make it super convenient to filter

for AI images even based on which model

they were generated in and whether their

original prompt is available or not out

of the 400 million stock images

currently on Adobe stock 10.1% are AI

generated contributors have to indicate

if their image was created using AI

tools and whether people or property in

the images are fictional here is one of

their guidelines for generative AI

content don't submit content created

using prompts containing other artist

names or created using prompts otherwise

intended to copy another artist the most

interesting thing is that adobe decided

to train their generative model Adobe

Firefly on a unique data set made up of

adobe stock IM images along with openly

licensed work and public domain content

where copyright has expired I stock by

gy images also introduced their own

generative AI solution in September 2023

partnering with Nvidia which similarly

to Adobe Firefly was exclusively trained

on their own stock Library these two

examples Adobe and gy pro that

generative AI models can work without

stealing billions of images and being

transparent about their training data

large stock sites like these were in a

unique position that they were sitting

on a gold mine from a deep learning

point of view they have a huge library

of mostly high quality and high

resolution images with descriptive

titles and tags added by the

contributors to make their images easier

to find at the moment when comparing

Adobe Firefly with mid Journey for

instance it's a clear wiin in most cases

for Mid Journey in terms of the quality

and realism of the output however Adobe

is undeniably the more ethical model out

of the two flip normals is another great

example of a digital Marketplace aimed

at 3D artists where the founders Morton

and Henning made it clear that

contributors can upload and sell a wide

variety of content as long as it's not

AI generated any AI art uploaded even if

given away for free will be taken down

relying mainly on generative AI to make

money is not going to last long even if

the Creator is going to try to hide the

fact that they used AI on one hand it is

unethical using mid journey is similar

to using torrent sites selling work

heavily relying on generated images is

even worse however it is like selling

stuff that you downloaded from torrent

sites the creators of AI tools are not

directly involved in the of plagiarism

they only provide the framework or

vehicle for it on the other hand art

created with AI is super accessible

anyone can do it which means pretty much

everyone will use it eventually it

reminds me of Google and other search

engines when they first came out in the

mid '90s and using them to find answers

to your questions was like magic it

seemed like a unique skill for a while

but then it became the norm and now it

is as normal to use a search engine as

breeding having amazing cameras on our

mobile devices didn't automatically turn

all of us into professional

photographers also randomly taking

thousands of photos every day may still

not result in any outstanding images

generating images videos music 3D

objects and everything else that will be

possible in the future will quickly lose

its novelty and hype and the value of

generated art will plummet Jake Parker

calls generative AI a new and Incredibly

successful form of industrialized art

where quantity affordability and speed

is the priority over quality and skill

previous successful examples of

industrialized art are stock sites

Fiverr and social media platforms like

Pinterest Instagram Tik Tok the

companies behind social platforms are

baiting users with pretty pictures and

videos harvesting their attention and

selling it to advertisers they are

serving up compact thumbnail size art on

a conveyor belt what's worse is that

these social platforms are already

saturated with AI images and soon AI

videos will follow turning them into a

Minefield if you're looking for

authentic content compared to these

other examples of industrialized art

generative AI takes things to a whole

new level it is a peak example of

consumerism it's like the the fast food

version of art it's available to

everyone it's cheap but it is not good

for you while fast food is negatively

affecting your physical condition AI art

is affecting your mind and how you

perceive things that you see looking at

generated images is numbing your senses

and distorting your perception of

reality remember when a piece of art was

experienced by staring at it for several

minutes and appreciating all its details

and imp factions trying to learn its

meanings and Mysteries the thoughts and

emotions the artist was trying to

capture generative images on the other

hand serve our impatient Society

perfectly we are Highwire or trained to

think that looking at anything for more

than 2 seconds feels like a waste of

time and to be frank that is actually

true about most AI art what humans want

is like the shortcut like biologically

we don't want to do work we don't want

to do do hard things because well I mean

we burn more calories and if we don't

have to we don't want to do that if

there was a pill that overnight you

would get jacked you would go from being

a skinny or a fat guy to like actually

like Henri caval in The Witcher kind of

level and you would be like sweet now

that's done I don't have to spend 10

years of my life and 5 hours at a gym

per day like you would you would just be

there and it's almost like we have that

if you have a button to get jacked with

in art like what's the point of things

right the real appreciation of art comes

from valuing the process of the artist

not just the time and effort it took

them to create a specific piece of art

but their entire life experience

generative AI removes the process from

creation and makes people believe that

it offers a shortcut to acquire artistic

skills and experience the actual Act of

taking a great photo requires much less

time than creating a painting of the

same sub subject and it could also be

considered a shortcut to achieve the

same composition however taking great

photos still relies on a lot of skills

and practice compared to writing a

prompt imagine that you have an idea for

a painting and you describe it to your

friend you also tell her that you don't

think you have the time or skills to

create it but you would love to see

someone else painting it next time you

meet your friend shows you a painting

that closely resembles what you

described she tells you that your idea

inspired her and that she worked on it

for several days now would you consider

yourself to be the artist in this

situation this is the same relation

between a user and a generative AI tool

where the AI is the friend who does the

actual work based on your prompt another

way to think about this is to imagine a

game where you have two options in the

main menu to start the game or to see

the end of the game the first option

would require you to spend dozens of

hours of playtime to reach the ending if

you decide to choose the second option

you could save all that time and effort

but would the ending mean anything to

you then not only you wouldn't fully

understand what's happening but you

wouldn't have any emotional ties to any

of the characters or events is AI art

not just a new type of digital art which

when it first became available was

considered to be a shortcut to creating

art and was looked down on by

traditional artists most people had the

misconception of digital art being

somewhat similar to generative images

where you just push a few buttons and a

great image pops out but compared to AI

art digital art is made by hand and

requires a lot of skill to do well most

AI artists already look at traditional

and digital artists as dinosaurs and

saying generative AI is just another

tool and those who won't embrace it will

fall behind but most of them don't

realize is that without an AI tool they

wouldn't be able to even come close to

their generated images while pre-i

artists would be able to make amazing

art pretty much with any tool you put in

their hands some some people who have

who have been into generative AI when

they saw the mid Journey artist database

they stopped using it right away because

they realized what the implications were

and they understood that it wasn't their

skill it was other people's skills and

they just

they they understood the the moral

implications right other

people just turned a blind eye and kept

going because it was too painful to

understand what was happening and they

just wanted to keep spiking their

dopamine and they wanted to stay in

these um communities without thinking

too critically right again this is

generalizing but a lot of people who are

using these image models never knew who

any of us were before which means that

their interest in art wasn't that deep

mhm it never went that deep to find out

who Kim jongi was you know who Greg

rowski was who John Singer Sergeant was

like none of that right there was no

real interest this just became junk food

yeah it's and it's it's interest based

on dopamine accessibility and

addiction right and it's causing harm

here is another little experiment for

you let's find a few great examples of

generated images from mid journey and

compare the complexity and

sophistication of prompts used to

generate them is there a correlation

between how detailed The Prompt is and

the quality of the generated image in

most cases there isn't which proves that

there is a huge factor of luck and

Randomness involved in generative AI art

democratization of art is an interesting

but flowed concept most AI artists who

who believe in this sentiment however

would still prefer to own copyrights for

their generated images they don't mind

exploiting other artist work but they

want their work to be marketable and

profit from it there is a big divide

between the AI artists and non- AI

artists or as jav Lopez founder of

magnific AI likes to call them the anti-

aai cluster the art community has always

been welcoming and helpful to anyone

interested to get started or gain skills

I hate to see this unjustified hatred

towards traditional and digital artists

as if they were holding on to their

secrets and acting as Gatekeepers

thinking that finally now with

generative AI anyone can become an

artist a lot of people look at this

movement as a revolution where the

snobby artists are finally getting what

they deserve while in reality nothing

can be further from the truth at the

moment the main reason why big Brands

and companies are careful about adopting

generated images and videos in their

communication is that AI art cannot be

copyrighted the US copyright office

noted that image generators produce

images in an unpredictable way and thus

cannot be considered creative or

inventive typing in a prompt and having

generative AI create the image we don't

think that output is copyrightable by

itself because we think that the um last

step of expression is being done by the

AI not you you're typing your prompt

you're like you know pink bear riding a

bicycle the AI is choosing in the first

instance what kind of bear the shade of

pink all the things that are supposed to

be the expression that the artist is

supposed to have in order to get a

copyright so we think that just typing

in a prompt is probably not going to

create a copyrighted blpr the copyright

law is quite clear that human authorship

is required for copyright adobe's

content authenticity initiative and

content credentials were created to

enable creators to add extra information

about themselves and their creative

process directly to their content in the

form of a new kind of temper evident

metadata this could be a solution for

proving the ratio between the work

achieved by Ai and the human artist but

we will have to see how widely it will

be accepted and how effective it will

become if you're a creative professional

you're never satisfied with what comes

out of one of these gen of AI um models

because it's not exactly what you wanted

right you're always going to make it

whatever your vision is and it's it's we

we referred this the first step in the

creative process and all the other steps

are going to be catable the opinion on

generative AI is extremely divisive and

I hope if you ended up watching this

video this far you can see why most

artists agree that they would consider

using AI if the training data was clear

of copyright violations using AI is

similar to tapping into the collective

imagination of humankind and it can be

useful for ideation for for the first

time you can brainstorm on your own and

get surprising ideas you may have never

thought of creativity is fueled by the

unexpected and AI has no fear of failing

so it will often make surprising

mistakes that can lead to breakthroughs

and Brilliant new ideas Sam Harper said

that AI art makes glorious mistakes and

therefore makes glorious art he also

said that AI is in its early days and

it's making lots and lots of mistakes

these are the days of AI being a good

artist once it gets too good at doing

what it was designed for it will no

longer make mistakes one question that

I'm particularly interested in is

whether there is a way to make a

completely ethical generative AI tool

Adobe has an AI Ethics program for

instance following three guiding

principles accountability responsibility

and transparency which spell art by the

way but it can be argued that shoving

all of adobe stock into a training data

set was not all that ethical since the

contributors only found out about this

after their images were already used to

train Firefly it is also fairly easy to

find generated images with copyrighted

materials on famous IPS on Adobe stock

I'm pretty sure Disney and Nickelodean

would not be happy to find out that

adobe is licensing their intellectual

properties I am sure that adobe is doing

its best to filter these out but it's

not an easy task when the contributors

are trying to be tricky when naming

these assets what I hope will happen in

the near future is that the L will

Define the requirements for generative

AI data sets so what would an ideal

ethical legislation look like for

generative AI training data sets to be

100%

transparent only use work in a data set

by artists who gave their consent

royalties and compensation should be

paid out to artists who opted in

generated images need to be cataloged

and easy to track direct or exact style

mimicry should not be allowed perhaps

the combination of minimum two artist

styles could work or simply completely

ban the usage of names of creators

generated images can only be used for

reference purposes and fully AI

generated work should not be acceptable

as a finished concept the UK House of

Lords publication on generative AI

states that the government should

prioritize fairness and responsible

Innovation it must resolve disputes

definitely including through updated

legislation if needed Empower rights

holders to check if their data has been

used without permission and invest in

large highquality training data sets to

encourage Tech firms to use license

material this is a very promising

Direction and I hope it won't take long

before this gets more traction around

the world the real danger of generative

AI is the effect it's having on upand

cominging artists many talents may have

been lost already to discouragement by

generative AI many might ask the same

question why should I spend over 10,000

hours perfecting my skills and repping

my head around complex and confusing

things like Anatomy perspective and

color theory when I can just write a few

words and get dozens of beautiful

generated images I already assume that

every day I'm seeing several AI images

that I'm not clocking right the the

problem with AI is that you're only

noticing the ones that you notice yeah

right the ones that are of sufficient

quality you're just scrolling right past

them you know you're just accepting them

for what they are and um I have

seen groups of professional artists you

know in a sort of formalized way try to

see how much can they identify AI art

and you know who can get a perfect score

right and in a considerable group of

artists you know maybe 20 or something

like that there's only one person who

will clock them all you know there and

and most everybody will make some amount

of mistakes and not just one so I'm

deeply deeply concerned about this I

think this problem is only going to get

worse and it it's bad for the artists

and it's also bad for the atmosphere

around Art Online the the issue with

stuff like this is that once there's

enough of this online and I think we're

basically already there you'll just

start assuming most stuff is AI and when

that happens we're on a slippery slope

that I really don't know how we come

back from once that happens that's a

very bad position to put new artists

into and artists who are trying to break

in and get popular it's horrible to

think that I would be less inclined to

recognize them as a a great talent just

because I'm like well where did you come

from and if you came out of nowhere

you're probably AI you know you don't

have a natural progression where have

you been it's like that is a horrible

atmosphere to have around art sharing

and the the and and social media I I

think that's only going to make us more

insular like you said it's going to make

us more inclined to only can only

interact with the people that we already

follow you know it's going to be much

harder to on board someone new as a

favorite artist or a great artist you

know think about the young people right

now I get so many comments and messages

from people telling me how they feel

hopeless there are literally kids out

there art babies who are in art schools

going through a 4-year program to

hopefully work in the industry and they

feel like there's no hope for them if

you're an artist with a voice with a

platform you don't have to do it for

yourself do it for the younger

generation there are so many people

young people who are counting on you I

don't know about you guys but I want a

future where there's an option to pursue

the passion of art as a career I want a

future where the kids want to draw want

to create want to make things that mean

something to them Stan preno or proo in

the draftsman podcast said drawing is a

visual language it's a form of

communication and just because a robot

can speak doesn't mean we should stop

speaking generative AI models are

devouring art in order to come close to

the greatness created by real artists

but they will never have a soul or

feelings to express but I do think that

also at some point if not already uh a

lot of people and a lot of companies are

coming to the realization that human

skill matters uh creativity really does

matter you know a lot of book covers a

lot of Publishers a lot of companies

creative companies are already

instituting no AI policies how long does

it take for AI to become the the sort of

the bottom line the minimal version of

what we can get and and if you want

something that is really surprising and

really thought-provoking and challenging

you'll still need to come back to humans

and at that point I don't know how many

human artists will be still thriving but

I do think at some point Human Art is

going to be value higher than AI are if

not already that that AI is always going

to be second rate it's always going to

be the cheap version when you can't hire

someone with real skill and creativity

you go back on Mid journey and you get

this you know uh something that you can

have for 50 cents um and so there's

going to be this constant race of how do

we tell the difference how do we

identify real human skill versus

something that just came out of a you

know pixel predictor um so that will be

interesting to see how it develops and

and what kind of tools come up with to

to to tell that difference to identify

something as you know true human

creativity versus uh a copy one of the

best things that happened recently on

YouTube that demonstrates the beauty and

importance of discipline and hard work

is the amazing art journey of one of

YouTube's biggest influencers PewDiePie

Felix decided he was going to improve

his drawing skills and Drew at first for

30 days then continued his challenge

till he reached 100 days his progress

passion and dedication is so inspiring

and sets a great example for those

discouraged from learning to draw some

companies like procreate are taking a

firm standpoint that they are not going

to introduce anything to do with

generative AI to their products they are

a small team passionate about art and

creation and they always value their

users over making profit their product

procreate has been the number one

bestseller app on the app store for iPad

for over 5 years and now their latest

app procreate dreams became a close

second on the bestseller chart they are

a prime example of a tech company who

believes in providing people tools to

express Express themselves easily but

they realize that AI is not about making

the process of creation easier it is

about automating or replacing the

process itself we're putting the whole

creative future of humanity at risk here

because another issue with this is not

just that uh like you let's say you want

to be an illustrator and now you go

straight to doing this kind of stuff

you're you're skipping the important

steps of learning about the art

fundamentals it's not just art

fundamentals as you know learning

perspective or learning construction and

color theory it's it's like you go

through a journey when you're learning

these things like you you're not just

going to school and like you don't you

don't enter the school as as the same

person as you leave the school as

because there is just an emotional

Journey there what does it mean to even

be creative or to express yourself that

is what what's at stake when you're

dealing with technology like this

generative AI is here to stay there is

no stopping or slowing down this

technology should you use generative AI

for your creative work I will let you

decide that but whatever you do I

encourage you to use this technology

carefully I am hopeful that generative

AI will evolve and become more ethical

in the future but until then we have to

be very selective in how we use it thank

you for watching don't forget to share

your opinion in the comments and I hope

to see you in the next

one