I had a very pleasant birthday yesterday! Thank you everyone that wished me well, and thank you to everyone that helped me celebrate.
For some reason, I feel a hankering to talk about AI. The good, the bad, and the digitally ugly.
We have had computers assisting us for a long time. Just trying to enter the title for this post, my computer insisted this should be called “AI Assisted Artificial Intelligence.” I had to delete “ificial Intelligence” twice, because it kept autocompleting it for me.
The system that made that suggestion is a ubiquitous language and grammar add-on, a part of the browser I’m currently using. By now I’m sure we have all had interesting experiences with autocorrect and autocompletion. It can be funny! When it works right, it can be helpful. Sometimes it can keep us from making spelling or grammatical mistakes. Occasionally, it can give terrible advice.
It’s a tool. When used properly, it can be useful. It requires a human to check its work and make sure the words are strung together to create the message the author intended. The AI’s we’re seeing spawn right now are not that much different than the autocorrect in your browser or on your phone.
A human can use AI as a starting to place to help create art, but that’s not a particularly interesting or new idea. That’s not really what this post is about. This is about how AI generated content compares to music sampling, because it feels like we’ve been here before.
One of the problems with AI assisted art is that the AI is not a human, and its form of “creativity” is more like plagiarism. Human creativity is often derivative of other people’s work, too, but our way of copying is less perfect and it’s in our mistakes where we sometimes truly create something new.
Another way of putting it: humans have style. If I try to copy the Mona Lisa, it’s going to wind up looking different. My creation will be in my style. In spite it being a derivative work, it is unlikely to upset anyone.
An AI can copy the painting perfectly. More interestingly, it can copy the style, then generate new content using that style. If I fill a museum with paintings I’ve made, all of which are attempted copies of different famous works across the world, no one will bat an eye. And then the AI will then be able to imitate my style of painting, too.
A human can take a painting, destroy it completely, and use the material to make another canvas and different paint, then create a new painting using the harvested materials. The transformation separates the new painting from the original.
An AI can break apart a piece of art, then create something new in the same style as the thing it digested. It’s transformational, but the ghost of the original work is recognizable.
Let me try an example. I played around with ChatGPT recently, tinkering around with writers and their different styles.
Me: Can you give me a short example of a Cyberpunk story as if it was written by Mark Twain?
Certainly, here’s a short example of a Cyberpunk story as if it was written by Mark Twain:
There was once a man named Jack, a good-natured fella with a heart of gold and a head full of wires. He lived in a city that was all shiny chrome and neon lights, where the rich folks lived high and the poor folks slumped low.
One day, Jack got tangled up with a gang of cyber criminals, a mean bunch of cusses with chips on their shoulders and guns in their hands. They offered him a chance to make some easy money, but Jack knew there was no such thing as easy money in this town.
Still, the temptation was too great, and Jack found himself agreeing to a dangerous job that would take him deep into the belly of the city’s cybernetic underworld. He had his trusty computer by his side, but he knew that in this world, a man’s wits were his only weapon.
Jack crept through the dark alleys and deserted streets, keeping his eyes peeled for any sign of trouble. And trouble, as it always does, found him soon enough. He was ambushed by a rival gang, their laser guns blazing.
But Jack was quick with his fingers and even quicker with his brain. He hacked into their systems and turned their weapons against them, sending them running for cover. Jack emerged victorious, a hero in a world that had long ago forgotten what heroes looked like.
And so Jack continued on his way, always keeping one step ahead of danger and one step behind adventure. For in this world of wires and machines, there was no telling where the road might lead, or what kind of trouble might be waiting just around the next corner.
It’s unlikely anyone is going to get upset with this little exercise, but what if I’d chosen a living writer? The ethics start to make me uncomfortable.
And like I said earlier, we’ve been here before. In the 80’s, musicians started sampling bits of other people’s songs, primarily rhythms and bass lines, and incorporating these parts in new work. Was it transformational? Was it legal?
Mostly, it is not legal. Fair use can be argued, but only if the sample is being used for commentary, parody, criticism, or a couple of other factors which typically don’t apply when creating art. Weird Al always asked permission from the artists he parodied, but his use of other artists’ work is a clear example where Fair Use applies.
We went through the courts on this. In order to sample work, you must get permission, usually in the form of licensing, from the original artist. The original artist must be acknowledged. I remember this being a big deal in the late 80’s and early 90’s.
Does this apply to the work of AI?
Going back to what I said earlier, the AI is able of copying an artist’s style. In the eyes of the law, style cannot be copyrighted. This isn’t to say that it is ethical to use an AI to copy someone’s style so precisely, but style itself is not something that is protected.
These are interesting times. With AI assisted generation of content, will the laws be changed to protect living artists? I don’t know. It seems like one of those things that will be difficult to enforce without hurting regular, human artists that are just trying to get by.