My encounters with generative AI

I’ve been following with great interest—and some worry—all the latest developments around artificial intelligence (AI) applications. After all, I took an elective class on AI in grad school decades ago. ChatGPT and Bing are being used to write news articles, college essays, resumes, and speeches. Midjourney and DALL-E are being used to generate imagery. These are only the best-known; they have many competitors, all racing to achieve the holy grail of producing results that are indistinguishable from the best human efforts.

Such disruptive high-tech innovation seems so distant from us artists working in traditional media, but I’ve already been personally affected by it:

• The de Young Museum in San Francisco holds a very competitive triennial (every three years) exhibition open to artists in all media in the nine Bay Area counties. For 2023, they added a category for AI-generated art. This caused quite a stir in the Bay Area art community; many artists are against AI being acknowledged as an art medium. Some suggested they would boycott by not entering. But that wouldn’t help, it just means the traditional artists who don’t enter would get less public exposure and the AI artists who do enter would get more.
• The Triton Museum of Art in Santa Clara, CA holds a statewide 2D competition annually. For 2023, a clause was added to the prospectus forbidding AI-generated art to be included in any way, with the threat of having the art removed from the show if it was discovered the artist used AI. They were praised by many artists for taking a stand against it.

These two stands on generative AI couldn’t be much further apart! And then there was the demo artist at the monthly meeting of a local art club I attended: his introduction that was read aloud was really something, gushing about him and his work, and it didn’t sound like something he could’ve written. He confessed later that he had an AI write it.

Those who support the use of generative AI for imagery like to point out that photography wasn’t accepted as an art medium for a long time, and the art world finally came around. I assert that photography is different: a camera, directly manipulated by the photographer, captures the world through a lens. AI creates a scene that doesn’t exist at all. What the best fantasy and visual effects artists conjure with a lot of imagination, skill, and hours of labor, an AI does in seconds from a few prompts based on the deep knowledge of imagery it’s been trained on. If you really want to compare photography to generative AI, it’s more like asking a skilled photographer to take a photo for you based on a few basic instructions and then taking credit for it. But the result is still their photo, not yours!

With an AI, anyone with no art skills whatsoever can toss out a few phrases and get back an image that could be mistaken for a master’s. And they take credit for it, even though they didn’t create it. How is a professional artist supposed to feel about that?

And then there’s the issue of how these AIs are “trained”. They are given volumes of images by every artist, sculptor, photographer, etc. so that one can ask for, say, “a painting in the style of Edward Hopper.” The “trainers” did not ask permission from any of the artists to do this. And there is a potential copyright issue around the creation of derivative work.

The whole thing is truly a Pandora’s box. It’s going to make the next several years “interesting.” I know I’m not going to convince anyone who likes the idea of using AI to generate imagery and passing it off as their own that it’s wrong.

If you haven’t guessed by now, here’s how I feel about it all: as a retired software engineer I admire the research and work that has gone into the breakthrough technology, but as an artist I think a lot more thought and discussion needs to happen before it’s granted wide acceptance. Probably even the passage of a few laws. Sadly, I think the horse has already left the barn.