Unethical or inevitable? Artists question where to draw the line on AI-generated art

A Halifax photographer using ChatGPT to transform her images into illustrations is opening up a conversation about the purpose of art in the age of artificial intelligence.

Lauren Thomander recently began selling a series of prints on her Etsy page inspired by the illustrated covers of The New Yorker magazine.

But when they were reposted on Halifax Noise, a popular Instagram feed with more than 80,000 followers, the response was mixed. 

Some praised the work and eagerly followed the link to her shop to purchase a print, but others, including many illustrators themselves, were quick to criticize the idea.

"People are kind of saying that AI trains itself with other people's work, and that it's not my work anymore," she said. "I don't see it that way because it is my picture that you can see there." 

In Thomander's view, using ChatGPT to make an illustration is no different than using software to edit photos. 

"Whether you're maybe changing the background of a photo, removing somebody that's in the way of your subject, or even putting a filter over it, I feel like that's all kind of in the same boat as AI," she said.

'The process is totally unethical' says illustrator

Illustrator and printmaker Alex MacAskill strongly disagrees. The print may have originated from Thomander's picture, but she's a photographer and not an illustrator, he says. 

"I think if you flip the situation on the other end and take illustrations you want to convert into photography, I can't help but feel that the person on the other end of the equation would feel similar to how illustrators feel in this particular situation," said MacAskill, who owns and operates Midnight Oil Print and Design House in Halifax.

He said Thomander could have realized her idea instead through collaborating with another artist.

"To replace that with AI … and to remove that person from the process is totally unethical," he said.

Large-language models, like those used to build ChatGPT or Google's Gemini or Anthropic's Claude, are trained by mining large data sets that include the creative output of many thousands of artists and writers, often without their consent.

When you ask ChatGPT a question, for instance, it's using that data to answer the query, which is why it's able so quickly to translate Thomander's work into an illustration that wouldn't look out of place on the cover of The New Yorker.

Photography was once seen as threat

It's no secret that many previous innovations in the world of technology have been received skeptically in the art world.

It wasn't that long ago that photography was viewed as an existential threat to painting, says Ray Cronin, a curator at the Beaverbrook Art Gallery in Fredericton.

"When photography was introduced in the 19th century, people were convinced it was going to kill painting," Cronin said.

Even more recently, when Mary Pratt, a renowned Canadian artist who was known for making lifelike paintings of everyday objects, began painting from photographs in the late 1960s, there were some people who thought she wasn't playing by the rules.

"Her parents thought she was cheating because she wasn't inventing the images herself," said Cronin, who wrote a book about Pratt's life and work.

Pratt would select an image from her collection of photos, maybe one she took herself. From there, she would project the image onto a blank canvas and trace the outline before starting to paint by hand.

It was all part of what Cronin says was her desire to "make the human eye look at the world the way the camera looks at the world." 

"The skill to do that is immense," he said " It's not something that could just be fed into some kind of algorithm."

When Pratt died in 2018, this was unquestionably true. The matter has become more complicated in the age of readily available generative AI.

A simple prompt on ChatGPT asking to "create a painting with apples in the style of Mary Pratt" will churn out an image that might, to a less discerning viewer, look just like the real thing. 

But this ignores a key quality of Pratt's work, Cronin said. Although photography played a role in the work, Pratt's paintings are special because they were made by a real person using real materials.

"Everybody knows the experience of seeing something, whether it's a photograph or a painting or a sculpture, that you can't just walk by. You have to actually look at it. And the longer you look at it, the more you see," he said.

Even still, Cronin is not ready to write off AI altogether. 

"I think AI is a tool and tools are only as good as the people who use them. And they're only as interesting as what's made with them." 

However, he hasn't found much art made with AI particularly interesting. 

"I'd rather see somebody with their mistakes and failures and experiments … than the kind of perfection AI always seems to create," he said. 

Imperfection makes art interesting, says Cronin

Art, in his view, isn't perfection. It's the absence of it. In Pratt's 1994 painting Glassy Apples, that much is clear.

The painting, based on a photograph she'd taken of a crystal bowl filled with apples that had been a wedding present, was made as her marriage to painter Christopher Pratt was falling apart. 

"The thing she always pointed out about that work that was most important to her is that there's a chip in the rim of the crystal bowl," said Cronin. 

When she took the photo in advance of the painting, she turned the bowl so that a chip was visible.

"She's showing that there's a crack in everything. It's breaking," he said. "That's not something you'd get from asking AI to make a bowl of apples in Mary Pratt's style."

Future unknown

For both MacAskill and Thomander, how AI will affect art in the future is an open question.

Beyond the ethical implications of using AI, MacAskill worries about what may happen if we lose sight of why art is valuable in the first place. 

"I think to value the convenience of creating something over the humanity behind it is really just doing a disservice to all of us," he said. 

But Thomander believes that as AI continues to develop as a technology, it will be prudent to get acquainted with how it works.

"It's better to kind of get on board with it, otherwise you might just be left behind in the dust a little bit."