AI generated art for writers
Is it a threat or a useful tool?
A couple of weeks ago I read an interesting article about the nascent state of AI art. Here it is:
It’s a subject that has since stuck in my brain and refused to go away, largely because I seem unable to settle my own thoughts on it. But as a writer, I can already see ways in which I could make use of the technology. In no particular order, then, here’s what my brain is doing:
This is clever
I don’t really yet understand how AI art works. My layman’s brain is assuming that it’s taking a huge wealth of source material and using it in some sort of machine learning neural net Cyberdyne system to generate new images, based on contextual input and assumptions about meaning.
Regardless, I find that a) it’s often super weird and b) it’s often surprisingly evocative. It’s also unexpectedly organic and malleable, in a way my 1980s brain doesn’t expect from a computer.
Take a look a this:
It’s odd. It has aspects of a photobash about it. The lizard from Amazing Spider-Man seems to be on the left. As a concept image it also sparks ideas about an archaeological dig on an alien planet. It raises intriguing questions about what is happening, in much the same way as a good piece of conecpt art.
A while back I stumbled upon Nvidia’s Canvas app, which is free to download, and which can be used to generate super-fast landscapes:
The results don’t currently get near to a proper artist working on a scene. But they do get pretty close to some of my amateur holiday photographs. Here’s another example:
It’s fascinating, and potentially useful as a starting point or for background use, and presumably is only going to get better.
I never enjoy this question, but I suppose we’ve got to go there:
BUT IS IT ART
I have no idea. Having had decades of “are computer games art?”, I’ve reluctant to go down this route. I feel unqualified to answer, and also uninterested in whatever the answer might be.
A more pertintent question for me is about the role of the artist. Does it matter if AI art was not specifically painted/drawn/crafted by a human? (there’s a related question of whether the creator of the AI system counts as ‘the artist’, or the person supplying the query prompt)
Purely in terms of the final image, I don’t know if it matters. If someone didn’t tell me, in a lot of cases I probably wouldn’t know if an image was created by a human by hand, or using an AI generator (in some less successful instances it’s extremely obvious, of course). However, there is a less tangible but important aspect: the knowledge that there is a person behind the created item.
I’m very aware that this may be a generational thing, and I am probably going to end up sounding like an old man. However, I visited the London Comic-Con last week and it was a an absolute joy to tour the artists’ alley and meet the actual human beings behind the created works. Quality of the material aside, whether I even liked it or not, knowing that there’s a person directly responsible does add something to the experience. A deliberate intent. When I watch a TV show produced by J Michael Straczynski, or read a comic by Brian K Vaughan or Kieron Gillen, it does matter that they created it. Whether or not I ever actually do meet them in person isn’t the important bit: it’s that it’s a possibility that makes it valuable. Even with long-dead creators, the idea that a person made a thing hundreds of years ago is enticing, and lends it historical weight.
I spent a couple of years producing a podcast about writing, talking to a different writer each week. The idea that you could have an intelligent conversation about the creation of a piece of art is a big part of what makes me interested in the art in the first place.
Ultimately, given two identically well-crafted artistic objects, one AI-generated and one made by a human, and told which is which, I would always opt for the human creation - for no better reason than I am also a human.
It’s less about the quality of the art, and more about a perceived, hard-to-define empathic connection to the creator. No matter how good an AI generator is, it’s hard to imagine having that link.
But what about games
Rather getting in the way of my last point is the existence of procedural generation in computer games. A decade+ back, there was a debate about whether procedurally generated assets and levels could be as good as a hand-crafted level put together by a designer. The general assumption was that no, it wasn’t possible. Procedural generation could be clever, but it would never be as satisfying.
I feel like that debate ended quite a while back. Hand-crafted levels are great. Procgen levels can also be great. Heat Signature’s spaceships, characters and stories are generated but feel tangible and engaging, albeit within the design confines of a very specific game. Minecraft is…well, Minecraft - a genius blend of procedurally generated and hand-crafted. No Man’s Sky is a similarly fascinating example. Invisible inc. has procedurally generated levels which feel like tightly crafted scenes from a Mission Impossible film.
Here’s me talking about Heat Signature, back when I used to make gaming videos:
Minecraft is an interesting example for this discussion, in fact. The worlds generated by the game are not hand-crafted, though they are driven by human-created code. But where the game gets especially interesting is in how players build on top of the procedurally generated landscape. I suspect that is where AI art is going to end up, at least in the short term: as starting points and inspiration for human artists to then further develop.
Hold on, is it ethical?
As I mentioned, I’m not yet fully cognisant of the technical aspects of how AI art works. Which means I have no idea of the accuracy of this tweet:
It doesn’t seem too far fetched to imagine that AI art relies upon huge datasets to function. That’s how machine learning tends to derive meaning - or the illusion of meaning. It seems a similar debate to the repurposing of journalistic material by Google and other aggregate newsfeeds, or the way Google generally uses other people’s content to create worth in its own search services. The difference being that AI art is specifically claiming to be generative, rather than simply representative.
On the other hand, human creators constantly use the work or other artists as inspiration. The books I read, the art I see, the movies I watch, all filter through into my own work. Humans build on what’s come before, that’s always been the case. We absorb, apply our own unique perspective, and create something of our own. To play devil’s advocate, is AI art doing anything especially different, conceptually?
I don’t feel informed enough yet to properly comment, but it’s a debate that’s going to get a lot more attention in the coming years.
Writers can definitely use it
My own selfish perspective definitely comes into play around this point. I write a couple of newsletters each week - this one, which focuses on writing and creativity generally, and then the Friday installment of my fiction. I create my own illustrations whenever possible, though sometimes have to fall back on a quick Unsplash stock image. It’s OK, but I wish I had more time to work on my own illustrations, and book covers.
The idea of being able to quickly generate images through an AI generator is extremely appealing. It would be more unique and evocative than a stock image, and faster than creating something myself. For Tales from the Triverse in particular, it could be a fantastic tool for conceptualising some of the creatures and places in the book.
It’s not just me, either: creating strong imagery for use on blogs and promotional materials is a real challenge for a lot of small organisations, especially non-profits. AI art could be very useful to fill in some of the gaps.
Sure, when it comes to something important where you can wrangle an actual budget it’d be preferable to hire an actual artist. But it’s those situations where that’s not an option that AI art could be an intriguing new avenue to explore.
I do entirely sympathise with this, though:
I hope that won’t be the case. Anyone who can afford an illustrator will hopefully still see the benefits of working with an actual human. Those scenarios where an illustrator can’t be budgeted for anyway - such as the images I create for this newsletter - is where I think AI art can really pick up the slack.
It’s probably ‘just’ a tool, right?
I’m very much an amateur illustrator. I love drawing, but am also very aware that I’m not particularly good and have a lot to learn. The idea of using AI art as inspiration and jumping-off points really appeals:
I can imagine generating a ton of AI art from specific prompts, in order to help me work on book cover concepts. Rather than using the generated image as the final work, though, it’d instead be a starting point: an intermediary prompt for me to take and then apply my own human layers on top.
What I don’t yet have is direct access to one of these systems. If I’m able to play with one in the near future I’ll certainly update my thoughts - and it might even start influencing the visual side of these newsletters. I’ll then be able to focus my thoughts somewhat, rather than relying on the opinions of others.
After all, this has all happened before and will happen again.
Many thanks to Matt Plummer for sharing some of his AI generated images, which were created using MidJourneyAi.
In other news, don’t miss this science fiction ebook giveaway in which you can find a collected version of Tales from the Triverse among lots of other intriguing titles: The June Sci-Fi Giveaway.