#10 Artist versus AI
How has art evolved over time? How did technological disruptions shape art movements in the past? How to deal with the issues of copyright and consent related to AI-generated art?
The beauty of a work of art is grounded in the context of the times and circumstances in which it is made. With changing context, what we recognize as art evolves too. What remains constant is the artist’s need for expression and their intent. While AI art has the potential to disrupt art, and questions about copyright and consent remain unsolved, artists are here to stay.
The painting on the left in the above frame, Girl with the Pearl Earring, is one of the most recognized pieces of artwork in the world. It was made by Johannes Vermeer in 1665, a period that is known as the golden age of Dutch painting. The painting has an aura, a sense of being alive, even though it has deteriorated and changed over the past 350 years.
The painting on the right is an AI generated artwork. Hague’s Mauritshuis, where the original painting resides, had organized a competition called ‘My Girl with a Pearl’ inviting artists to reimagine the painting, while the original painting was touring for a Vermeer exhibition at another museum. One of the top 5 winners was the Midjourney AI created piece called ‘A Girl With Glowing Earring’. This has been a controversial selection and many have outraged over putting a machine generated piece on the same level as human created artwork. Read the details here.
I leave it to the judgement of the reader if the AI art is doing any justice to the place that it is occupying.
The AI Explosion
In the past year, we have seen an explosion of generative AI tools ranging from text to art generators to a host of chatbots that can imitate human-like responses. I assume almost everyone reading this must have tried some variant of these, so I will not go into listing them here. In this post, I will dive into the topic of AI-generated art and the many issues that surround it.
Last year, I had done a puliyabaazi on this topic when OpenAI had just released the Dall-E model for everyone to try. I had tried to answer one fundamental question— how did art evolve when there were technological disruptions in the past? Taking stalk of this, I had made two predictions:
We will see a new genre of AI artwork being created and we will ultimately become okay with calling it “art”.
Artists who make art today will have to focus on the point they are making with their artwork. Artists will have to sharpen their intent because that is a domain still in the artist’s control.
After a year, I do a review of the AI art scene. Do I stand vindicated? Let’s find out.
What kind of AI art is being created today?
I did a quick search on Twitter using #aiart. Can you guess what is the most repeated subject of AI-generated art? Take a hint - in the early days, what was the most common thing people searched on the internet? Yes, and yes. You are thinking in the right direction— Salacious pictures of scantily clothed women with impossible body curves. I must say that this finding affirms my belief that the more things change, the more they remain the same. Now, I will not blame AI for this. A score of digital and non-digital artwork objectifying women has remained an issue of debate even before AI came to the scene. AI has merely given tools to those who did not have the talent to bring their fantasies to life.
Some genuine Use Cases
But this is not all. I found some genuinely interesting use cases too. Look at the below snapshot taken from twitter. (P.S. I do not know these people, just using them for example). Looks like someone made an interesting picture of their teacher using AI art. The teacher seems to be finding it creative enough and is touched by the expression. I would say that qualifies as art.
It seems like a thing many people would like to try out. Sure enough, there are people on Fiverr ready to provide similar services. As someone who has tried getting work done using this platform in the past, I would temper my expectations. We will talk about the copyright and ownership related issues with these images later. But it is worth noting that graphic artists have already started using AI to earn money in one form or another. There are a number of people offering to train a LoRA model for you, create AI generated images with your products, or create AI avatars of you.
A big chunk of these AI images fall into the anime, sci-fi, and fantasy genres, and so all of these look very similar-ish. These genres were already quite popular with digital artists, so it’s not surprising that AI tools are now being used to create numerous spin-offs of the same styles.
Can AI-generated art be considered art?
To answer this, let’s delve into the question of how art has evolved with changing times, tools, and technologies. I acknowledge that this quick history I narrate is limited to Western art movements because although I love and admire many of our Indian and indigenous art forms, I am not very well-versed in their histories.
Renaissance Art
What we call art is continuously evolving. About 500 years ago, it was the State who was the storyteller. Hence, a majority of art was being made under the patronage of the church or the temple to tell the stories that the State wanted to tell. So, there was the Renaissance art movement where Michelangelo was painting the ceiling of the Sistine Chapel and da Vinci was painting the mural The Last Supper (1498 CE).
Tools were limited and in absence of any other technology, the challenge for the artist was imitating life. How lifelike can a painting or a sculpture be? This is the period when we get the anatomically perfect sculpture of David by Michelangelo(1504 CE) where every little muscle is carved masterfully.
Baroque Art (early 17th century to mid 18th century)
This changed with the Baroque period from the early 17th century to mid 18th century. Mind you - baroque is a style of art not only in painting and sculpture, but also in music and dance. The term baroque wasn’t a flattering one. The sentiment was that this was a corrupted form of art. This was also the time when artists were experimenting with new and vibrant colours to play on the light and contrasts in paintings.
See the two examples below which have a wonderful play of light, shadows, and depth that was missing in the painting of earlier periods.
There was also a change in the subject of these arts. Earlier, a painting typically featured either a mythological character or a King or a wealthy man. In the baroque period, common people such as milkmaids started getting featured as subjects of artwork. Our Girl with the Pearl Earring also falls under this era.
Photography & Impressionism
Let’s fast forward by a couple of centuries and come to 1822 when photography was invented. Now imagine yourself as a portrait artist who is earning a decent living by drawing life-like portraits of wealthy people. Now, you have competition from a gadget that can do the same in much less time without much talent. So what do we see happen after that? Did artists stop making artwork or portraits after the mid 1800’s?
What we see is the emergence of a new movement in art called impressionism. This again was a derogatory term used by a critic of that art style.
With the advent of photography, we can say that the incentives of artists to paint like real life reduced. There was a tool that could do it better than you. So, you would want to differentiate yourself in some other way. Impressionism can be seen as a response to that line of thinking where the focus shifted away from copying life to capturing the feel of life. The above painting may not look like a realistic sunset, but it feels like a moody sunset.
This thought process continued to evolve during the 20th century and you see more and more abstract forms of art where the whole point is either to communicate a certain feeling or to make a statement. Lots of people continued to make beautiful photorealistic art too, but what really catches our attention today are art pieces that make a point, irrespective of the level of craft involved in doing that. An example of this is a controversial piece called the Comedian which was essentially a banana taped to a wall that got sold for 120000 USD.
New art forms made possible by technology
When photography was first invented, it was not considered an art form. But today, we accept photography as art. Not only that, but photography evolved into videography and that evolved into a completely new art medium called movies. We can say that photography and movies democratised art. Today, anyone can have a selfie taken, while earlier it was a privilege reserved for those who had the money to get their portrait painted. Objectively, more people are creating art and enjoying art today than 500 years ago. Technology has not only expanded what is called art, but also expanded the number of people who can create art as well as enjoy art.
What happened to portrait artists?
While photographs became common, rich and famous people continued to have themselves painted by talented artists. John Singer Sargent was one such prolific portrait painter during those days. Even today, many portrait artists or even sketchers make simple, affordable hand-made portraits that people cherish for their personalized feel.
So then, is everything ok with AI-generated art?
Nope. There are two issues - copyright and consent.
Copyright is probably the easier one to solve, but right now the policies seem very unclear. The privacy and copyright terms and conditions vary across different platforms.
If you are on free trial on Midjourney, you do not have the copyright. By default whatever you create is viewable and remixable to others. There is a Stealth mode you can pay for, but their policy wordings on this is very non-committal.
The answer to the question “Will images created on Midjourney in paid mode be shared with others?” seems to be “kinda.. sorta..we will try our best not to”. The hunch is that this caveat means that your images will still be used by Midjourney to train its model. If you are working on a project that you do not wish to keep in the public domain, you would be wary of using Midjourney for it.
Here’re couple of lines from Midjourney’s terms and conditions:
If You purchased the Stealth feature as part of Your “Pro” or "Mega" subscription or through the previously available add-on, we agree to make best efforts not to publish any Assets You make in any situation where you have engaged stealth mode in the Services.
Please be aware that any image You make in a shared or open space such as a Discord chatroom, is viewable by anyone in that chatroom, regardless of whether Stealth mode is engaged.
I must also add here that trying to use Midjourney reminded me of my Unix command prompt days. An AI generator that takes command prompt is quite ironic. One more instance of “the more the things change, the more they remain the same”?
The bigger issue is about consent
Whether to consider AI art as genuine or not can be a subjective question. However, objectively we know that these models have been trained on images that were created by real artists, and this has been done without their consent.
If you are an artist that has put out any piece of art on the internet, be sure that it has been used for training. You can go here to run a search against the LAION-5B dataset that is used by Stable Diffusion. All of this art is the basis on which these AI models are spewing AI-art in the style of numerous living and dead artists. It’s one thing to generate art in the style of Van Gogh, it’s another matter to generate art in the style of a living artist. Especially, when you use an artist's name in the text prompt, you are acknowledging that you are using something of theirs. It is only fair that the artist be paid for what is being acknowledged as theirs by the thousands of prompts that use their work and name to create AI art. There should be a mechanism for the artist to “opt-out” from their art being used for training and generation purposes.
But these images were publicly available and anyone can get inspired by them!
I don’t think that there is merit to this argument. We cannot use any publicly viewable image on an artist’s website or portfolio for commercial use. Even training these models seems like something beyond fair use and should require prior consent. These AI models are monetizing their output. It is a clear commercial purpose and should require prior consent as well as compensation. Artists are literally being ripped off, and for the time being they have no recourse other than outraging on the internet.
Authors and artists have also filed suits against AI companies OpenAI, Stability AI, Midjourney, and Deviant Art. Details here.
Sarah Andersen, a noted cartoonist who is a party in the suit, writes:
“My drawings are a reflection of my soul. What happens when artificial intelligence— and anyone with access to it— can replicate them?”
I agree with her sentiment. We can also look at the question from an incentives framework. When an artist’s work is unfairly used, and such a use is normalized, how will other artists have confidence and incentives to create artwork in public domain that may be prone to similar misuse. Their incentives will shift to creating art only for big art studios and movies which will at least ensure that the art is attributed to them and that a decent compensation is provided for it.
I think that this is a change that AI companies will have to accommodate for. Artists should have the ability to opt out from the training dataset. We have a copyright framework that works well for books, music and art in general. Why should AI companies be exempt from it? The AI models can give a compensation based on the frequency with which an artist’s name and art style is invoked. This may attract more artists to create newer styles that work well for AI art. That could be a potential win-win.
A massive reskilling spree
Artists are not alone in this. It is clear that we will see a massive reskilling exercise for all white collar workers to be trained on AI enabled tools. Is it going to be a bit like the computerization phase in the 80s and the 90s? Or will it be more disruptive than that? Only time will tell.
The above illustration from Harvard Business Review magazine, Sep-Oct 2023 edition about reskilling is quite apt. It is a pretty clever illustration by human artist Noma Bar and validates my second point— artists will have to focus on the intent because even in the world of AI-generated art, the intent still belongs to the artist.
And that’s all for today! If you have enjoyed reading this, please consider subscribing to this newsletter. You can also leave your comments - your words of encouragement will mean a lot to me.
Thanks,
Khyati
P.S: The past week on Puliyabaazi, we got an opportunity to speak to Pratap Bhanu Mehta on India’s democracy. If you can understand Hindi, it would be a conversation worth your time. Sharing the link below: