This post is part of Lifehacker’s “Living With AI” series: We investigate the current state of AI, walk through how it can be useful (and how it can’t), and evaluate where this revolutionary tech is heading next. Read more here.
AI-generated art isn’t a concept: It’s here. Thanks to numerous tools with simple and approachable interfaces, anyone can hop on their computer can start generating whatever image ideas pop into their minds. However, as more people have started experimenting with these tools, serious ethical and legal issues have cropped up, and just about everyone online seems to have an opinion on this divisive technology.
As part of our series on living with AI, we put together this guide to demystify how AI art tools work, explain the controversies around them, and show how they impact everyone, from professional artists to curious casuals.
Where and how to make AI art
Table of Contents
Before we get too far in the weeds on the tech and ethics of AI art, let’s quickly overview the tools themselves.
There are many AI art generators out there, but the major players would be Midjourney, Stable Diffusion, Bing’s AI image creator, DALLE-2, Craiyon, and Dream. All of these tools are accessible via the web or desktop, and Dream also has a mobile app.
Midjourney is considered the most powerful, but it requires a subscription starting at $10 per month. Midjourney also requires a Discord account, since it operates entirely through a dedicate Discord chat server. That means you’re working alongside other users, and all your images are posted publicly on Midjourney’s web gallery, unless you pay $60 per month for the “stealth mode” feature.
DALLE-2 also generates high-quality images, but new users must pay at least $15 to sign up to use it. While your session and creations are private, it uses a confusing token system where each word in your prompt costs a certain amount of money, and you need to pay extra if you go beyond your token limit.
Bing’s AI generator and Stable Diffusion are entirely free and let you make as many images as you want, but image generation takes longer, especially if the servers are busy. Craiyon is also free, but subject to longer generation times, and image quality is lower. Lastly, Dream offers a decent middle ground between quality and price, letting you make one image at a time for free, or up to four with a paid account starting at $10 per month.
In terms of approachability, Bing’s AI generator is by far the best option if you’re just curious about these tools. It’s free, and you can access it from the Bing search homepage or in the Edge browser.
That said, despite the differences in quality and and interface, these tools all work the same way: You type a prompt into a text box describing the image you want to see, press enter, then wait a few moments for the AI to generate the picture based on your description.
The quality of the final product will depend on which tool you’re using and how detailed your prompt is. Some tools, like Midjourney and Stable Diffusion, have guides on coming up with better prompts, and even extra features that can help the AI get closer to your intended result. But even with these extra steps, the process only takes a few moments, and every tool is simple to start using.
How do these tools learn how to draw? And why are there too many fingers?
The images you get from these tools can be impressive, but it’s not because the software actually knows how to draw.
As I pointed out earlier this year, calling these products “AI” is a misnomer. Unlike the common conception of artificial intelligence as seen in science fiction media, these tools are not alive, sentient, or aware in any way, and they do not reason or learn. This is true of both text-based chatbots like ChatGPT and generative art tools like Midjourney. In the simplest terms, they work like your phone’s predictive text, pulling from a list of possible solutions to your prompt. When it comes to generative art tools specifically, the tool simply searches for images that match the keywords or descriptions in your prompt, then mashes the elements together.
This is entirely different from the actual process of drawing—in fact, the AI never “draws” anything at all, which is why these tools are notorious for “not knowing how to draw hands.”
As video game designer Doc Burford explains, “If I tell a machine ‘show me Nic Cage dressed as Superman,’ the machine may have images tagged with ‘Nic Cage,’ and it may have images tagged ‘Superman,’ but where the thinking mind of an actual intelligence will put those ideas together and fill in the blank spots with things it knows—like an artist who has also memorized human anatomy—the AI is still gonna give me an imperfect S-Shield on Superman’s chest, it’s gonna mess up the fingers.”
Ethical and legal concerns of AI art
These tools are easy to use and can oftentimes spit back compelling results—discounting the occasional extra digits and wonky faces—but there are major ethical concerns around making and distributing AI-generated art that go beyond quality and accuracy.
The main issue with generative AI art tools is they’re built on the backs of uncredited, unpaid artists whose art is used without consent. Every image you generate only exists because of the artists it’s copying from, even if those works aren’t copyrighted. Some AI evangelists like to claim these tools work “just like the human brain” and that “human artists are inspired by or reference other artists the same way,” but this is untrue for multiple reasons.
First, there is no being or mind in an AI, and therefore no memory, no intention, and no skill. Stable Diffusion isn’t “learning” how to draw or taking inspiration from another piece of art: It’s just an algorithm that searches and auto-fills data in the ways it’s programmed to. Humans, on the other hand, think, feel, and act with intention. Their works come from their memorized skill and lived experience. Even using another person’s art as a reference or inspiration is a deliberate choice informed by the artist’s goals.
To help explain the distinction, I reached out to Nicholas Kole, an illustrator and character designer who works with major film and video game studios like Disney, Activision, and Dreamworks. “The work I do, as a concept artist and illustrator, begins with digging deeply into the context of each project,” he says. “I ask pointed questions, tease out ideas about worldbuilding, story, gameplay [and] the process from start to finish is extremely specific, bespoke, and tailored to the precise needs of my colleagues and clients. Every single cufflink, belt buckle, prop, and motif—the art we make is the careful work of thoughtful design, done lovingly and with attention to detail.”
“The intrusion of an algorithm into that process, that doesn’t care for context, that doesn’t know whether people have 5 or 17 fingers, that mashes up visual guesswork based on stolen data, and functions essentially like a million monkeys at a million typewriters is anathema to me.”
Kole says AI art “flies in the face of everything I stand for creatively, and everything I’ve wanted to do with my life’s work. It’s an insult to the reason I make and engage with art—I want to see thoughtful human craft that is expressive, and express myself with thoughtful human craft.” A quick glance through portfolio sites like Art Station shows Kole is not alone in those sentiments, with many professional artists taking a hard stand against AI art.
This hardline stance isn’t just for ideological or aesthetic reasons, either. AI automatization poses a threat to job security for many industries—it’s why unionized actors, screenwriters, Google contractors, and even airline workers are currently on strike around America. The threat to working artists is just as real.
AI art also poses a risk to the companies that employ these artists. There have already been major legal battles over AI art infringing on copyrighted materials, with the current outcomes favoring the original artists. As such, many companies outright ban the use of AI art and reject any applications from artists with AI-generated works in their portfolios to avoid any copyright issues.
Are there ethical uses of AI art?
Despite the ethical and legal issues, some argue there is a place for these tools, and that they can even be helpful to professional artists. In an interview with Kotaku, visual artist RJ Palmers says artists could use AI to “come up with loose compositions, color patterns, lighting, etc,” for example, and that the tools “can all be very cool for getting inspiration.”
Similarly, author and animator Scott Sullivan argues in his blog that AI is helpful for ideation and iteration while brainstorming, and that “it’s all about the artist’s intention and how they use the tool.”
But while AI art is contentious among professional artists, casual users may wonder whether any of this matters for the layperson or hobbyist that just wants to play with them once in a while. And, sure, AI tools could be used as toys, but it’s important to note that’s not how the creators of these products treat them.
Almost all AI art generators are commercial products in some way. Some are paid products like Midjourney or DALLE-2, while free services like Bing’s AI image generator earn Microsoft money through ad revenue. Some are also used as “proof of concept” examples to entice commercial clients to pay for the more powerful version of the tool.
In all cases, the people that make these tools earn money off the work of the artists whose work is used to create the image you’re generated, even if it’s just for fun. As Kole explains, “The generative system can’t function without the stolen lifes’ work of countless passionate people just like me. They each brought their lived experiences, opinions, fixations, and points of view to their bodies of work, only now to have them smashed together inexpertly and touted as original art.” Even if you don’t share or sell images you make, many of these tools keep a public record of all generated content that other users can download and distribute.
Given all these concerns, it’s hard to recommend AI art creators, even if the intent to use them is innocent. Nevertheless, these tools are here, and unless some future regulations force them to change, we can’t stop folks from giving them a try. But, if you do, please keep in mind the legal and ethical issues associated with making and sharing AI art, think twice about sharing it, and never claim an AI-generated image as your own work.