AI-generated art is a really, really big deal

Here’s a bold statement: I’ve never seen anything as incredible in my lifetime as what is happening right now with AI-generated art.

I still like to consider myself pretty young, but in my few decades here I’ve witnessed the internet go from a nerdy novelty to the connective fabric of society, computer graphics go from the original Nintendo Entertainment System to photorealistic feature films, and all of the world’s knowledge organized, connected, and shoved into our pockets. All of these things are mind-blowing when you look at them now through a long-term lens, and many steps along the way were pretty damn impressive even in the moment.

But what is happening right now with artificial intelligence is moving in units of days, not years. Every week there is something new and heretofore unimaginable on this rapidly unfolding frontier.

AI-generated art is what is what it says on the tin: images created by a piece of sophisticated software. There are a number of competing projects at the vanguard of this space, but they all more-or-less work the same way: an artificial intelligence was trained on thousands and thousands of reference images, learning what it meant to for something to look like an “apple” or a “cat,” or what it meant when something was an “oil painting” or “GoPro footage.”

After all that training and programming, the software can respond to a prompt from a user, like “an oil painting of a cat eating an apple,” and spit out something that looks like this:

That’s cool already. But the real fun stuff comes when you get a little weird with it, like “the Empire State Building as a muppet,” or "an iPhone drawn by Pablo Picasso."

If this all stopped here, with funky images generated from silly strings of words and styles, it would already be incredible. But this is just the beginning. Over the past few months, we’ve seen a deluge of innovations and experiments:

• Cosmopolitan created the first ever AI-generated magazine cover art
• Pieces generated by software are already winning first place in art competitions
• A designer held a virtual fashion show with pieces swapping live on models
• A fake TikTok creator pranked his audience into thinking he was AI-generated, still showing off some impressive tools in the process
• An urban planner is using AI-renders to showcase what cities would look like people-friendly streets

We’re just getting started, and already we’re grappling with ethical concerns around AI-art. A creepy early use case of these tools, like many things in the history of technology, involved nonconsensual pornographic images known as “deep fakes.” Artists with distinctive styles have complained that these platforms enable a form of high-tech plagiarism. Related, the debate is just beginning about who owns what when the intellectual property is the product of some code instead of a human creator.

This pace of innovation in this space is only accelerating – and we can begin to see where it is going. With the right tools and talent, is already possible to create fictional digital media that is so realistic that the average person cannot tell it apart from truth. Right now, those tools are hard to use and the talent is in short supply. But soon, very soon, those barriers will dissolve, and as we reach a world where anything can be created by anybody, those good-intentioned and those with more nefarious aims, our ability to trust what we see and hear will disappear along with it.

The defining labor and economic issue in this century will be what happens when more and more of our lives get automated – from driverless cars to algorithmic bankers. The defining social and cultural issue of this period just might be what happens when that same animating technical spark challenges what we consider art and ultimately, what is truth.

Go play with it yourself and see

There are a number of projects doing fascinating stuff at the moment. The most popular one you'll see shared on Twitter and elsewhere is Craiyon, formally known as DALL-E Mini – a rapidly developed tool inspired by the field's biggest player, OpenAI's DALL-E 2, which is more resource intensive and requires an invite. Similar platform MidJourney is open for beta, and can be used via a Discord instance that also lets you see everybody else's generations in real-time. And to round out the biggest players, Stable Diffusion has had a lot of recent buzz and is available for download to run on your own computer here, but for those less technical you can try it online here.

‍Each of these tools has its own idiosyncrasies and you might need to play around a little bit to find the type of prompts that will give you the best results. But once you do, you'll be blown away by something, to paraphrase from F. Scott Fitzgerald, commensurate with our capacity for wonder.