Using AI Ethically

 

Midjourney-generated image using the promts “pint of pale beer, pub, atmospheric, shadowy, people, smoky, wood paneling, edward hopper”

Welcome to AI week! For the next few days, perhaps not constrained to this calendar week, I will be considering the myriad ways AI may change our lives. The technology has many upsides and possibilities, but some of them are harmful. In today's post, I consider the question: how do we use AI ethically?
 

The great thing about AI is that it uses the entire internet as grist for crafting words and images. The terrible thing about AI is that it uses the entire internet. AI has access to all our collected wisdom—and all our bigotry, confusion, and errors. The power and the danger of AI are two sides of the same coin. With all that power comes the opportunity for misuse. It’s fun to ask AI to compose a poem about your partner washing the dishes in the style of Dr. Seuss, or to ask Midjourney to create an image of your dog in the style of Pablo Picasso. The results may not be amazing, but the mash-up is fun.

Yet particularly with imagery, this power can do a lot of damage. The photo at the top of this post was created by Midjourney, and I used Edward Hopper as a prompt. It’s appropriately Hopperesque; the algorithm successfully captures the distant look on the man’s face the artist was famous for. But what if I hadn’t used Hopper, who died in 1967, and instead plugged in Em Sauter’s name as a prompt? I’m not going to do it, because the result would likely loose a fake-Sauter into the world. She’s a working illustrator, and depends on sales of her work and commissions for her income. If everyone can get a “Sauter” illustration by asking the computer to steal her style, what happens to Em?

 
 
 
 

This is not a hypothetical—it’s happening right now.

Last year, Kelly McKernan, a 36-year-old artist from Nashville who uses watercolor and acrylic gouache to create original illustrations for books, comics, and games, entered their name into the website Have I Been Trained. That’s when they learned that some of their artwork was used to train Stable Diffusion, the free AI model that lets anyone generate professional-quality images with a simple text prompt…. Anybody who used Stable Diffusion, McKernan realized, could now generate artwork in McKernan’s style simply by typing in their name. And at no point had anyone approached them to seek consent or offer compensation.

Kelly, along with two other artists, is suing three AI image generators, Midjourney, Stable Diffusion, and DeviantArt. They believe, and I agree with them completely, that this is theft. Whether or not the law also agrees is another matter—and I have my doubts that the current courts are going to look favorably on artists who are up against big companies. (Check out Sarah’s piece at the New York Times for an example of the infringement.) To make this all the more serious, it doesn’t look like the creators of these engines thought much about the havoc they might be wreaking. Midjourney CEO David Holz described building his algorithm, something like a thief describing how he plundered a jewelry store:

“It’s just a big scrape of the Internet. We use the open data sets that are published and train across those. And I’d say that’s something that 100% of people do. We weren’t picky. The science is really evolving quickly in terms of how much data you really need, versus the quality of the model… There isn’t really a way to get a hundred million images and know where they’re coming from. It would be cool if images had metadata embedded in them about the copyright owner or something. But that's not a thing.”

I love Midjourney and I plan to use it for imagery on the blog going forward, but it’s worth reading that interview to hear just how clueless and reckless Holz is about his creation. I had flashbacks of starry-eyed Mark Zuckerberg talking about how social media was going to bring us all together and make the world a happy place—right before it did the opposite. The problem with Zuckerberg’s philosophy—move fast and break things—is that you end up in a world of rubble.

 

Using AI Ethically

Whatever the courts decide, people are using AI to create images and text now. I’m not actually too worried about writers getting ripped off. Aside from doggerel, it’s hard to get ChatGPT to use a writer’s voice effectively. Even when I asked it to write like Hemingway, Cormac McCarthy, and Salman Rushdie—writers with famously strong voices—it came up with pretty lame stuff. Most writers have far more subtle voices, so ripping them off is going to be next to impossible. Text generators can be used to spin out horrible stuff, though. It’s worth noting that since AI learns, it picks up human pathology along with its knowledge. So while theft isn’t a big danger, people can use these algorithms for malign purposes. Don’t use AI for evil!

Generated by Midjourney

With artists, the issue is more serious. Generators like Midjourney are already being used for this kind of theft. When I asked Em about this, she said, “It’s scary. I’m 100% not ok with it.”

So first and foremost, never use the name of a living artist or photographer as a prompt. Artist have a vision that includes not just visual form, but meaning. Art is a communication, and artists have strong and intentional points of view. In fact, this is what Em worries about. She referenced Sarah Anderson’s editorial and how anyone could use her art to make bigoted images: “People who make beer education that’s not accurate and ruin the credibility I’ve created.” Any time someone uses the artist’s style for their own purpose, they steal the artist’s voice.

Not stealing an artist’s work is a low bar. A more subtle danger is the way in which AI images are about to flood the internet—swamping the work of real artists who take brush to paper (or stylus to screen). And I relate to this—it was the conundrum writers faced when the internet arrived. It was so easy and cheap to create and distribute text that the value of creating it plummeted. Writers didn’t need the internet to steal from them; merely flooding the internet with low-grade content was damage enough.

It won’t be long before people can make credible AI music. Text generation may soon automate some or much of the work human journalists do. And we’ve talked about illustrations. (Futurists imagine many other applications that are further out, but will replace human creativity.) One important further step then is to clearly identify all synthetic content for what it is. When we make art on Midjourney, we shouldn’t try to pass it off as original work. It’s not original—algorithms are creating the art, not the Midjourney user. It’s good to let people see how ubiquitous these prices are becoming—and also makes original art stand out.

When we use AI, in any form, we should label it as such. It’s ethical because we shouldn’t be taking credit for AI, and because it protects the work of real people. In the coming years and decades, we may grow to value the human hand. Now is the time when we create good habits that distinguish between human and AI, and preserve the work of original creators.