AI: The New Artists?

Tuesday, December 12, 2023

AILAS' AI Odyssey/Art/AI: The New Artists?

AI: The New Artists?

The fact that AI is great at reading, writing, or crunching numbers is obvious to anyone who's seen it try. And honestly, that's what we've always expected AI to be great at. But what no one saw coming, is just how creative AI would get, and how quickly. Almost immediately after OpenAI released Chat-GPT and changed the world of AI forever, a new phenomenon arose: AI programs grabbed their pencils and paintbrushes, and started making art.

And man, have they gotten good at it.

The unrest in online art communities has been ongoing ever since. New art students were surprised to find their abysmal odds at gainful employment after graduation had just sunken even lower. And recently even Hollywood has gone on strike, with both actors and script writers fearing for their livelihoods.

 Clearly something's going on in the world of AI, and it's ruffled some feathers. But where did it all begin?

How did AI even figure out art?

Simply put: the same way it figured out everything else. AI learns by looking at examples of a given thing, until it understands what that thing is.

To start simply
: if you show an AI 100 images of a horse, then it will understand what a horse is and will be able to create a 101th. The 101th will be based on the previous 100.

Now, show it a hundred examples of the art style of chiaroscuro, which is about the contrast between light and dark. It can be animals, people, landscales, or objects... Like us humans, the AI is incredible at pattern recognition, and thus it will figure out what it must understand as "Chiaroscuro".

From there, "A horse in the style of chiaroscuro" is a simple matter for the AI; it already knows exactly what makes up those two, and can combine them seamlessly.

The same goes for "the style of Thomas Cole/Picasso/Vincent van Gogh". Show it enough examples, and the AI will figure it out. 

Outpacing human artists

If you've ever tried to find out whether there's the talent of the next Picasso within you, then you've found out also how hard it is to learn to draw anything.
You can't just draw a horse; you need to figure out the shape, the relative size, the position, pose, texture, and much more, just to draw a decent picture. Artists spend years practicing their art before you even see them. If you want to commission one, they need to fit your commission into their schedule. They need to be in a creative mood at the time. And they can only work in the styles that they've already practiced countless times.

Saint George on Judgment Day, in chiaroscuro.

In comparison, an AI program can observe countless images and just figure it all out. Once it's been well-trained, it can create anything, in any style, and imitate any artist. Most of all, it can perform all that within ten seconds. Any time of any day. Free of charge.

And as for quality? AI-generated art has won awards, to the anger of human artists. Their outrage is understandable, but think of it this way: that means a qualified panel of judges has sat down to examine dozens of beautiful artworks, only to decide, knowingly or not, that the AI-generated one was the best of them all.

And it probably took five minutes to make.

Creating art professionally

We've already explored where AI-generated art came from, and how it can create beautiful image at a fraction of the time and effort that it would take a human artist. Thinking pragmatically, as businesses are bound to do, this opened up the door to marketing unlike any seen before. I for one, have personally seen the proprietor of a small food delivery service create product images of his recipes by describing them to an AI, which he then put in his brochure. I've also seen a friend save big through replacing an interior designer, by creating designs for his new home until he knew exactly what he wanted, without paying anything for the idea.

Don't take my word for it; take a look at these:

The argument against AI-generated art

It's safe to say that these developments haven't been without their fair share of criticism. We've already talked about various types of artists being angry at what they perceive as a threat to their job security. But they're not the only ones, and theirs is not the only reason, to be critical of AI-generated imagery. Let's take a look at the major arguments.

1. AI-generated art isn't original.
Without doubt the most common argument, and certainly not an inconsiderable one: an AI cannot truly create anything new; it can only replicate and mix the things it has already seen.
And yes, this is true. Given the way the AI learns to create art, which is through learning by observation, it cannot truly imagine anything beyond that which it already knows. It follows that it cannot then create anything truly original.

On the other hand, those in favour of AI-generated art are quick to point out that human artists cannot create anything truly new, either. There is nothing new under the sun, and the human mind cannot express something that it hasn't already perceived in some way; or that is an amalgamation of things it has already seen. Examples of their rebuttals are "Try to come up with a new colour", or "Imagine a new animal, that isn't either a variant or a chimaera of ones that already exist."

Even when new styles are created by artists, they're not anything truly original, so the argument goes. We've looked at chiaroscuro, which is based on the mingling of light and dark. The style is beautiful, but it's nothing we haven't already experienced in the real world.
When Giger created the artistic style of biomechanica, he was quite literally combining two concepts which already existed. And when Lovecraft wanted to create the most outlandish thing possible: a series of unfathomable beings from outer space which defied human understanding, like his infamous Cthulhu, what he came up with to represent that was simply a mixture of several different animals. Something no human has ever seen, but a mixture of things that it has, combined into something new. As we've seen at the start of this blog, that's exactly the modus operandi of AI.

Both these artists were lauded for breaking new ground. Certainly they've done much more than what I described above, but these are their most famous creations. And since we call these creations original and groundbreaking, and the method used to create them was the same that AI programs use, it's hard to argue that the creations of AI are not or cannot be original and groundbreaking as well.

2. AI-generated art isn't real.
Another good argument. Earlier, I mentioned the proprietor of a small food-delivery service using AI-generated images to advertise his food, instead of just taking pictures of what his kitchen actually puts out. There may be a big difference between the flawless, succulent steak the AI imagines for him, and the lukewarm block of dogfood that he actually sells. AI has a tendency to idealize. But so does traditional product photography. Just think of the advertisements you see for fastfood, showing food that always looks much more appetizing than what they actually serve.

At the same time, one may easily respond that if authenticity is our worry, we need not think about AI any more than about the model-work we see being done for just about any advertisement, which looks something like this:

  • Select the most beautiful person out of a thousand (or more)
  • ​Have them adhere to a highly rigorous diet for months on end
  • Apply professional makeup
  • ​Put them in a studio under picture-perfect lighting
  • ​Shoot anywhere between fifty and hundreds of shots with a professional eye, and posing
  • ​Select the single best shot out of all those, then edit it, removing blemishes, changing the body and adding a filter

All this creates a truly beautiful picture. But is it authentic, when you pick the most beautiful person you can and then that's still not enough, and you have to add makeup and photo-editing? Product photography is done through a similar process, as well. If authenticity is the goal, then we may argue that creating something from scratch is no further removed from reality than taking something real and editing it until it becomes more beautiful, superficially, than it ever will be in real life.

3. AI-generated art isn't artistic.
The most poignant argument. And perhaps the most convincing one, as well. It is said that art captures the current zeitgeist before we understand it, or can put it into words. The point of "real art", whatever that might be exactly, is to capture the human experience. And then to communicate that to us in a way that nothing and no one else can. If it didn't, how else can you explain millions of people visiting the Vatican, or Michaelangelo's David inspiring people centuries after its creation? One thing that AI cannot do, is experience the world as a human.

In conclusion

AI-generated art has come, and it is here to stay. Whether that is truly a good or a bad thing, I leave up to you to consider. What we do know for certain, is that it will help a great many people still, while putting many others out of business. Most likely, the smartest thing to do is to use this power of AI for your own benefit if you can; but to always remain ethical about it.