The Grizzly in the Gallery: Redefining ‘Art’ in the Age of Generative AI
Is Generative AI Output ‘Art’?
This week, I was musing about whether the output of AI models can ever be considered ‘art.’ It’s a touchy subject (for artists in particular). I've worked in the creative industry long enough to remember that when we needed images for this or that, there were basically two paths to take: 1) hire a photographer or fine artist for custom commissioned work, or 2) purchase “stock” assets (either Royalty-Free or Licensed). For a time, there was much hand-wringing about how cheap and easily available royalty-free stock would basically drive the real artists out of business and back to painting tourist caricatures in Central Park.*
There were disruptions, sure, as artists tried to figure out what to do, but before long, new revenue streams were created. Who among us remembers grabbing our digital SLR and bolting together an ad-hoc stock photography business onto everything else we were doing? It seemed like things worked out okay, and a new equilibrium was reached. Then AI burst through the door with all the subtlety of a cocaine-addled grizzly bear wielding paint brushes. Just like that, we were gripped by existential dread all over again.
AI as Art Democratizer
Are assets created with Generative AI models valid artwork? My gut says yes, and here’s why. When I create artwork using Gen AI tools, I use my human wetware to write a prompt in highly descriptive language to tell the model what my creative vision for the image is. The AI then uses the data it has been trained on (more on that in a bit) to render my vision. Frequently, I need to devise additional prompts to tweak elements or edit something that didn’t match my original idea. Lather, rinse, re-render. Gen AI may be doing a lot more of the heavy lifting than, say, if I created an image from scratch using tools in Photoshop, but the final Gen AI image would not exist without the prompt I carefully constructed. That’s my creative input.
In a sense, Generative AI has become the great art democratizer: with AI, everyone has potentially become an artist, no formal training required. I’m not so elitist to think one needs formal training to be allowed in the clubhouse as a “Creative,” so don’t throw your clogs in the machine just yet.
It might be helpful to remember that artists have embraced all kinds of technology, from video and mechanics to VR and blockchain, seeking to push the boundaries of creative expression and placing technology into the role of partner. Consider:
- Nam June Paik: The founder of video art, Paik used TV monitors, radios, and robotics to construct sculptures, trying to humanize technology.
- Ian Cheng: Uses game engine tech to create live simulations that run autonomously, examining how technology impacts human evolution.
- Rachel Rossin: A multimedia artist combining painting and code, creating immersive VR installations that explore memory, physical, and virtual spaces.
- Manfred Mohr: A pioneer of generative art who has used computer algorithms to generate digital paintings and sculptures since the late 1960s.
What’s Fair Is Fair, Love and War Be Damned
The tough nut to crack is the issue of what Gen AI models are being trained on in order to bring my idea to life. Too often, AI models scoop up whatever they find on the internet and don’t lose any electric sleep over compensating whoever created it. (Props to Adobe, which uses licensed and public domain content to train its Firefly models). Early in the streaming music days, there was much concern about musicians not getting a royalty cut of all the streaming and how it might (did) cut into album sales. Although the payouts are very low, some form of resolution ultimately found its way to the streaming music world. I believe something similar should be put in place in the Brave New World of Generative AI. It’s only fair that LLM owners should have to compensate creators for the hard work that goes into creating art, thereby adding value to their models. Perhaps someday we can even expect LLM models and Gen AI tools to be stamped certified as “Creator Friendly” – indicating that creators were compensated fairly for their contributions to training the Machine. I’ll leave the messy details to someone else.
* With no prior knowledge of the field, my Dad once told me that when he heard I had decided to major in graphic design, he feared I would end up on the street drawing caricatures for tourists. Thankfully, it never came to that. I’m terrible at caricatures!
Ready to see how strategy-first design can scale your startup brand? Learn more about the power of purpose-driven identity.
Contact Us →