background

πŸš—πŸ–ŒοΈ Are we the old people yelling at an AI cloud? (Spoiler: no, we are not)


Images or email not loading properly? View it in your browser here.

Receive exactly what you are interested in

This newsletter goes over several topics.

You received this email because the following is one of your topics of interest:

  • Illustration
  • Photography
  • Stories and Experiences

​Manage your topics of interests here and receive exactly what you enjoy.


Today we will dive deep into something that's resonating with a lot of us: the use and abuse of generative AI.

We'll go over things like

  • What is (not) art?
  • Art being fundamentally human.
  • Why generative AI is theft.
  • How it differs from human inspiration.
  • How we might lose art altogether if we continue this route.

One topic, some explanations, a pinch of roasting and a blended choice of a gloomy or bright future ahead.

If this recipe sounds interesting, buckle up and enjoy the ride!


Chapter 1: What is (not) art?

Sorry, I am not here to define what is art. But what it is not.

AI-bros as non-artists don't understand that "an image" or "a song" are not art in and of themselves.

They are the output.

Saying that such output is "the art" is like saying that "cooking" is "the food".

This would mean that if you put takeaway, at home delivery or a supermarket packed meal "on your table", this would imply that you cooked.

No one would ever consider that cooking.

Chapter 2: Art and humans

Us human creators ENJOY creating.

The fact that AI is trying to "create art" instead of helping artists, shows AI-bros pushing this know NOTHING about creating or being artistic.

They are missing the point entirely.

AI should be used to help humans create more art

What we need is for AI to give humans more free time. It should take chores off our plate. People would be able to create and share more art.

Let people be more fulfilled and generate emotion in others through real art.

AI is trying to remove the human element from the fun part of art creation: art creating itself.

There are a ton of things involved in creating art.

Many are enjoyable, others not so much.

If AI "does" the fun part, what's left? The boring part that we actually dread? That's completely backwards.

You can't separate art from the human who makes it

Art can take many forms. But it has one thing in common: it is linked to a person and its relationship with, for a lack of better concept, "the universe".

It's a way of saying things that you can't put into words.

A way of sharing feelings that can't be shared otherwise.

An attempt to make people experience the world in the way that you do.

Art in its core is inseparable from empathy.

One enjoys seeing other people's art because it shows more than what they can say with words. You can see what they see, and more importantly, how they see it. What they like. How they feel.

This is why I think that…

Chapter 3: AI image generation is not art

Stop.

Stop calling "AI-Art" to "AI image generation".

It is not art.

Never was. Never will be.

Chapter 4: What AI image generation is: theft

Not to get very detailed in technical matters, you might have heard this one before.

AI uses something called LLMs (Large language models) and in a very simplified way, this is how it works:

  • A company creates something called a "model".
  • A model is basically a black box. You put something in and it spits something out. You don't know how it works on the inside.
  • The model, to be "effective", needs to be "trained" with a lot of data.
  • Model training is, again, in a very simplified way, showing it a ton of "inputs" and telling it "these are the expected outputs".
  • After training, the model can receive inputs it never received before, and "generate an output" based on "what it was trained on".
  • It's kind of like picking "a piece of each ingredient" (more on this analogy later) to generate such output.

Here is where theft comes in.

The part where the model is "trained" to be "effective" involves using existing art without getting consent or paying the artist. Basically stealing from them.

Have you ever gotten your video muted, demonetized or taken down due to a copyright claim?

That's someone exercising their rights on their art (music in this example). If you are profiting by using someone else's work, they need to agree to that (give consent), and be compensated fairly. As simple as that.

AI-bros feel they are above all this and that they "need a pass" or their "business model" will not work.

Well, cry me a river, you failed businessman.

AI models based on theft is by design

For more details, read this article from The Verge.

But the article says:

Nick Clegg, former UK deputy prime minister and former Meta executive, claimed a push for artist consent would β€œbasically kill” the AI industry.

This is not the first time I've read this. Let me get this straight.

"If they don't let you do it, your business does not work."

So maybe they are stealing, they are crappy businessmen and it's not much of a "business" model?

It's basically complaining because the police takes you to jail for robbing people on the street.

"My business model is robbing people on the street. Very profitable. But if they don't let me do it, my business model does not work! So please, let me do it!"

And the truth is, these things will really not work if they had to ethically source the material to train the models. There just isn't enough artists willing to give their work (free or paid) in order to train a system that might put them out of a job in the future.

Chapter 5: But what about inspiration?

Let me stop you right there.

In short: human inspiration and model training work vastly different. Leading to different results.

If you want a more real, "technical" explanation of why it's not the same, please refer to this thread by @svltart. You can also read this article by Karla Ortiz to go deeper into the subject matter (Follow them on Bluesky here and here).

The interpretation in my own words goes as follows.

What people think happens

You have this "magical entity" and give it some "input".

This can be words (a description) or even an image to "use the style of", or to "modify".

Black box stuff happens, the input gets "interpreted" and "processed".

You then think you get an original output.

How can it not be? It even gives you different results "each time" you ask for one.

What really happens

In reality, the part of "it gives different results each time" is false.

If you asked to it enough times, it's bound to repeat the result.

It might sound impossible, but let me explain.

Understanding "very high quantities"

Many of the misconceptions people have regarding these "ultra high capacity computing models" (call it LLMs, AI, Machine learning, etc.) is related to our inability to understand the dimensions of it. We are simply not used to doing so.

If you studied mathematics for a while, you might have encountered the concepts of "infinity" and how there are "bigger infinities than others".

At first, this sounds ridiculous.

Then it actually makes sense, but even after understanding how can this be possible (mostly by comparison), it stays weird. We struggle making sense of it, even if logically it does.

How it works: cooking edition

When an AI model is trained, there is an obscene amount of data that goes into it, gets processed, and another obscene amount of data goes out of it.

It's like having millions of ingredients, making every meal possible with them, and storing it. Then when someone asks you for a meal with certain ingredients, you pick the pre-made result off a shelf.

The key part of here is "quantity". Each ingredient can be used in different amounts. So, if you only have flour and water, you can have from 0 to 100% of each, and everything in between.

But you are bound to only create things that have flour and water. And that's the main limitation of these models. Once they have the ingredients (what they are trained on), they can only use that, and nothing else.

Add some olive oil

But you as a human, can think outside the box.

You can look at the flour and water, and think "what if I add a couple of spoons of olive oil?".

That, and the fact that humans can't replicate pixel by pixel someone else's art, is basically the difference between human inspiration and how AI models work.

You see things. You interpret them. Identify what they make you feel. The combination of inspiration sources plus your own experiences in life, makes you feel something different altogether. That's what you end up creating.

Bear in mind that when we talk about "ingredients" here, it's what the model is trained with. Not the prompt. You can change the prompt all you like, but the training/ingredients you have available stay the same and you have no influence on them.

Chapter 6: Non-artists, please hear me out

Artists already know most of this.

I want to raise awareness on non-artists.

People who are indifferent or "don't really mind" about gen AI usage.

We are starting to suffer "art loss"

The more gen AI is used:

  • Less work goes to real artists
  • Less people get into art
  • Less real art gets created

And this is a problem.

Encouraging what I call "art loss" will make our enjoyment of art worse in the future.

We will become more and more numb to art. We will start to ignore it just as "another AI crap that means nothing". Real art will get lost amongst the noise.

We, as humans, will lose.

And maybe that's the ultimate goal.

If we stop enjoying art, we have more time to spend looking at things they want us to look at, like ads.

Instant gratification, engineered just for you.

Getting you to only pay attention to what they want, because they know how to get you.

Chapter 7: But not all is lost

I have stuff to write. Things to draw. Experiences to photograph.

I want to create and share.

I will never stop doing art, because luckily I don't live from it. I do it just because I enjoy it SO much.

But I know a lot of people that do, and it will make me sad if they are not able to do it anymore because art is devaluated in a sea of crap art wannabe.

If you're an artist, don't give up.

If you're not, don't look the other way. You lived without AI generated images until now. Don't let those corporations win by using their AI.

Support real art and their artists.

I even encourage you to give art a try. It is rewarding and fulfilling. Sometimes frustrating.

But art makes you human. And no one will ever take that away from you.

Let's keep in touch!

That's it for today. If you enjoyed it, let me know! See you on the next one.

All the best,

Juanma from Creating Lightly

113 Cherry St #92768, Seattle, WA 98104-2205
​Unsubscribe Β· Preferences​

​Links to buy things may be affiliate (If I remember to add them lol). They don't cost you extra and it's a way to support me.

background

Subscribe to Creating Lightly email community