Does AI Create?
There’s a lot of debate over the ethics of AI taking away jobs, particularly “creative" jobs. I have been mulling over the question "does AI create"? It’s called “generative AI” not “creative AI”. So what’s the difference between generating and creating? The words we use are so important, and it’s imperative in this debate we are all using words in the same way, with the same definition.
Here’s the first issue: “create” has many definitions. Merriam-Webster has six:
Transitive Verb
to bring into existence: God created the heaven and the earth
to invest with a new form, office, or rank: She was created a lieutenant
to produce or bring about by a course of action or behavior: Her arrival created a terrible fuss
to produce through imaginitive skill: create a painting
Intransitive Verb
to make or bring into existence or something new: an artist who is good at imitating but not at creating
to set up a scoring opportunity in basketball: create off the dribble
Let’s get a quick refresher on transitive vs. intransitive verbs. A transitive verb requires a direct object to receive the action of the verb, typically answering the question of what or whom. An intransitive verb is not used with a direct object, typically answering the question of where, when, how, or how long.
To put this into practical terms, let’s use Andy Warhol'spainting and prints as an example. Andy Warhol used his hands to create an image of Marilyn Monroe, then used a printing press to create multiple copies of the image his hands created. Warhol "original prints” are now sold for thousands of dollars because, even though the prints are not the image directly created by Warhol’s hands, they are considered to be his creations. If I paint an image that is an exact copy of Andy Warhol’s Marilyn print, I am the creator of a replica that (should) require me to pay royalties to Warhol regardless of whether I used my hands or a machine to generate the replica. I may be the creator in the transitive definition but I did not create the image in the intransitive definition. What about the print my siblings and I had created for my dad 20 years ago that is a photo of us rendered in Andy Warhol’s style?
Creating new work based on past concepts is not new. Isn’t everyone learning from the works that came before and building upon it for their own work? Haven’t all the great masters built upon the work of artists that came before them? Don’t all great musicians get inspired by music that came before? So when AI is generating writing or images or sounds, prompted by a human with an idea, is the AI creating or the human? AI is a tool that human creators use to get their ideas out into the world.
There is an intellectual property question about the line between influence and usage, but it’s not a new problem. Everyone (or, everyone of a certain age…) remembers the infamous interview with Vanilla Ice about sampling “Under Pressure” that they added a note, so it’s different. [Side note: we can debate separately if he was really making a joke that was taken out of context.] We, the people, influence those laws by electing officials that represent our views and lobbying or protesting for our values. AI is vectorizing art and music in a way that we can mathematically measure the differences between works. Let’s use that math in our laws. How similar do two works need to be to considered imitating or sampling and require royalties?
Let’s also be honest about something: artists would not be mad about AI using their work to generate new works IF THEY GOT PAID FOR THE USAGE. How do we create systems that measure which works are used by the AI to generate something new, then pay those artists for the use. Today, artists are agreeing to exploitative terms of use for services they need to survive, especially independent artists. Giving Google the right to use any content uploaded to YouTube in its AI training models is exploitative. Artists are uploading their copyrighted work to YouTube, and then those copyrights are effectively nulled by the terms of use. This is what is meant by people‘s life work being stolen by AI. Not just that it’s being used to train the models, but it’s using copyrighted works without royalties. When music gets sampled, like "Under Pressure" in "Ice Ice Baby", the music industry found ways to make sure that the original artist got credit. How do we apply similar thinking to AI sampling? And not just for music, but for all creative arts: visual, text, and audio.
So what does it mean to create in the age of AI? Is creating different from generating? And how do we respect copyrights and intellectual property when AI companies are winning lawsuits with “fair use” arguments?