Don't be afraid of infinite beauty
Artificial Intelligence is capable of producing genuinely beautiful art – and that will be genuinely good for the world.
Part II of a three part series on the future of Artificial Intelligence
[Part I: How to think about machines that think]
This post focuses on how AI tools will change the nature of art and storytelling.
Inside this issue:
Will machines replace artists?
There’s no such thing as artificial art
Good artists copy, great artists steal
How to eat at the infinite buffet
Will machines replace artists?
The first time I wrote about artificial intelligence for Something Interesting was in December of 2022. Midjourney v4 had just been released and for the first time ever AI was making images that looked "real." That meant artificially generated images could seem photorealistic (more on the implications of that in the next post in our series) but it also meant for the first time AI could produce images that were beautiful, as opposed to just interesting.
There were still lots of visible artifacts (strange geometries, extra fingers, etc) but the improvements between v3 and v4 had clearly crossed a chasm. Suddenly even people with no artistic talent could produce aesthetically compelling, custom imagery practically for free and the technology continues to get better. For anyone who enjoys beautiful imagery, this is good news — beautiful imagery is cheaper and more abundant than we could ever have dreamed a few years ago.
For anyone who sells beautiful imagery, though, the situation is more complicated. At a minimum AI tools will transform their commercial reality — and many may find themselves entirely displaced. AI tools are likely to transform basically every job, but artists are especially sympathetic symbols of this cultural anxiety. They usually choose their careers for passion, which makes their job losses especially poignant and the prospect of new jobs cold comfort. And most importantly there is something essential and human about the act of making art. It feels sacred.
To meaningfully predict the impact of artificial intelligence on artists we need to be specific about what kind of art we mean. Some art is produced purely as an act of joy, such as a personal hobby or a gift to a loved one. That art was never commercial, so the existence of new commercial tools is not a threat. Amatuer musicians didn’t stop playing instruments when recorded music became cheaper, because playing music was never about efficiently making the desired sounds. Art that was produced for meaning will remain meaningful in a world where it is less expensive.
Anyone pursuing a skill for its own sake is unaffected by the rise of AI tools — just as home pianists weren’t obsoleted by player pianos. Anyone pursuing a creative vision on the other hand will suddenly find themselves unconstrained by the limit of their skill. They will be free to make more (and more ambitious) art than they could have on their own — and they will still be able to imbue those works with emotion and personal meaning:
Some art on the other hand is almost purely commercial — stock photography, for example, or corporate logo design. The artists working in those fields are seeking to create beautiful work, but not in service of their own meaning. They are ultimately seeking to fulfill the creative vision of their clients — and in many cases the clients may soon have the ability to fulfill their creative vision for themselves. AI tools create an abundance of technical skill — which means the jobs that consist mainly of renting technical skills out to others are probably going away.
Most artistic jobs are a blend of both creative vision and technical skill — but AI tools will affect those two aspects of artistic work very differently. Creative vision will remain scarce and expensive, but AI tools will make technical skill cheap and abundant. Machines won’t replace artists in general because there will still be artists operating the machines — but many jobs will be lost and the ones that remain will be radically (in some cases unrecognizably) transformed.
What that means for creative jobs as a category is difficult to say. Since AI will allow anyone to cheaply and easily create high quality art for themselves the level of creative talent required to sell your art to others will rise. Many artists will learn to their discomfort that although they thought of themselves as creatives, the market mainly valued their technical skills. Median and mediocre artists will likely stop being employable, in much the same way 'tavern musician' stopped being a viable profession by the time anyone could turn on a radio.
There’s no such thing as artificial art
A lot of people (particularly those who personally identify with the loss of this class of artistic jobs) employ various No True Scotsman fallacies to claim that AI assisted art can never be 'real' art. They’re making the same mistake so many people make of trying to ascribe agency to the tools rather than to the human using them. Claiming that the artists using AI tools aren’t really creating the art they make is like claiming a photographer isn’t really taking photographs, or like telling a musician they aren’t really playing music because the instrument is doing all the work. People have been making that joke since the 1700s!
"It’s easy to play any musical instrument: all you have to do is touch the right key at the right time and the instrument will play itself." — J.S. Bach
Our emotional connection to art is a not a function of the tools that were used to make it. People can connect emotionally with a capella singing and they can connect emotionally with electric guitar riffs and they can connect emotionally with the electronic beeping of a Tamagotchi keychain. The idea that people won’t emotionally connect with AI assisted art is pure Luddite copium. Even people who prefer unassisted human art for some domains will almost certainly make or consume AI art in other areas they are less personally interested in, in much the same way that people who enjoy marathons still mostly drive to them.
Good artists copy, great artists steal
A more subtle and interesting objection raised by AI skeptics is the complaint that because AI models were trained on images without artist permission they are in some sense an act of theft, or plagiarism. I am sympathetic to the people who raise this objection, but not to the objection itself. Copyright protection (and IP protection generally) is not a natural right — it is a legal construct designed by the state to serve our collective interests, not the interests of copyright holders.
Accusing AI companies / models of plagiarism is cathartic but it is naive to think that adding new copyright regulations would benefit small artists. The major beneficiary of expanded copyright law would be Disney. Creating rules that prevent AI models from learning without permission would result in crappy public models trained on public domain data and expensive corporate models trained on licensed data sets. The Disney™ AI image generator would be just as dystopian for freelance artists, it would just give all the benefits of AI art to Disney so they could rent some of it back to us at enormous profit. Not the win for human creativity people imagine.
We don’t need new copyright law to protect artists — we already have rules to protect artists from plagiarism. The reason these problems seem new is because people give too much agency to the AI tools and not enough to the humans who use them. If someone uses AI tools to plagiarize an artist, existing laws already apply. Inventing new laws to hold tools accountable for potential plagiarism doesn’t make sense.
What the people accusing AI of theft are really hoping to do is to extend the definition of plagiarism from "taking someone else’s art and presenting it as your own" to "using someone else’s art in a way they don’t like." There is already a rich body of case law governing what rights an artist has to dictate how others are inspired by or reference their original work. Fanfiction, parody, review/commentary and transformative art have created a lot of well established precedent for what kinds of "similar" we should allow in art and what kinds we should restrict. The existence of new tools for creating art does not imply the need to restrict new categories of art.
Ultimately the people who focus on creating red flag traffic laws and other obstacles to AI adoption are looking backwards. I think it is more interesting and valuable to look forward at what might lie ahead.
How to eat at the infinite buffet
I’ve mostly been focusing on AI generated images so far but it is easy enough to see that we are climbing similar exponential curves for video and audio production as well. I think we can reasonably assume that all (or nearly all) media will soon be AI assisted and that there will be a lot more of it. Given that American adults already consume more than 13 hours a day of media there probably isn’t going to be an equivalent growth in demand for media to consume. Instead, we should expect the competition over that pool of scarce attention to become incredibly fierce.1
Since better AI tools will make it harder to compete on production value alone, content producers will need to specialize in other ways. Just like the internet, AI media production will make viral hits even more viral and niche communities even more niche. The question is, in a world where anyone can tell a story skillfully, whose stories will we want to hear? If you could write the perfect story for yourself, would you want to read it? Or would you rather read an equally skillful story written for you by a loved one? Or by a public figure you admired?
There are other interesting questions as well. How will we feel about stories that never actually have to end? What if you could always request the next chapter, another season, one more movie? What if you could bring Sherlock Holmes and Harry Potter along with you to any storyline you wanted to experience? What if you and your loved ones could become characters in the story? How should we balance the desire to customize the story to an individual against following the artist’s creative vision or creating a shared audience experience?
Living in an age of abundant art will be like walking down an infinite buffet of delicious food and deciding what to do with your limited appetite. Do you try a bunch of different things? Or just fill your plate with the most delicious thing you can find? Do you eat the things that look the best to you or do you get the same things as your friend so you can eat them together?
The future of AI art is a future of abundance. It is the future where human creativity can finally break free from the limits of technical skill. Of all the ways that artificial intelligence will transform society art is by far the least alarming. If AI displaces artistic jobs it won’t be because there is less demand for art — it will be because more people can make art for themselves!
That might be a painful transition — but it will ultimately be a good thing.
Disclaimer: My wife is the founder and CEO of an AI powered interactive children’s storytelling startup (Wander.ly) that I do some of the art for and about which I am obviously quite biased. Definitely check it out!