Self-published Authors Worried about AI Artwork on Book Covers


Over the last six months we’ve seen rapid advancement in AI Image Generation and there is a lot of chat and controversy around the subject. So understandably a lot of authors are somewhat worried whether it’s a good idea or not to use AI art for book covers or pictures in an illustrated book. 

So I think it’s time to get my thoughts down and dispel some of the myths and tell you how it works and where you stand. 

So here you’ll find the facts.

Controversy #1: It’s All a Bit Dodgy, Run by Dodgy Disruptor Companies

Of course a lot of these services, such as Dall-E, Stable Diffusion, MidJourney, no one has ever heard of before AI Image Generation. And yes, these are the first movers in the field. But there’s a number of well-known names already on this too.

Microsoft, under its Bing banner has a product coming up. More info about that here.

ShutterStock, the place I’ve used for stock images for ten years, is getting in on the AI Image Generation act. They have a white-labelled version of Dall-E running on their website. Go look for yourself here

Big companies like these aren’t getting involved without it being 100% legit. It’s just that some of the lesser-known companies are the first movers in the space.

But if you want a deeper understanding of the subject, the controversies surround it, then sit back, get yourself a nice cup of tea and let me try and guide you through it. 

There are quite a few topics to cover. It might take a while but it will definitely leave you a little more informed on:

  • What your legal rights are, as an author.
  • How AI actually works.
  • Where the controversy lies.
  • What’s generally super interesting about it.
  • And finally, what the future holds.

Who I Am

If you don’t know who I am, I guess it’s worth pointing that out, to give you a little background. I’m James, the chap that runs the book cover design service. I’ve been a freelance graphic designer since the nineties. In fact, this is an important point, because I was designing over the internet in pre-broadband days, when you had to dial-up via a landline to get the internet. 

As in, I’ve always been ahead of the curve somewhat when it comes to technology. Because I love technology. I always have. In fact, another lifetime ago, I did a Computer Science degree at Brunel University. I dropped out in my final year back in ’94 to do more graphics stuff. I found drawing fun pictures more interesting than writing buggy code.

The point is that when some new technology comes along I’m a vociferous reader and thinker on the subject. Because I’m generally incredibly interested. And AI is no different. I’ve actually been reading on the subject for many years. 

I knew it was in the post. But AI Image Generation took me by surprise last year. 

But the fact that I’m a designer, and run a business, I have to be above board with everything. I can’t leave my backside out there flapping in the wind, so to speak. So this is something that I’ve thought about and looked at in thorough detail.

So let me try and explain how it works in the simplest way I can for you.

But before we get to the fun stuff about AI Image Generation, let’s get to one of controversies. One that just leaves me scratching my head. And that’s Copyright and Rights.

Controversy #1: You Can’t Copyright a Book Cover with AI Generated Images

The concept of copyright seems pretty straightforward to most authors. You can write a novel and copyright that novel, and own the copyright. Job done. 

When it comes to book covers the situation isn’t as clear cut.

Let me explain why.

A novel is made of elements called ‘words’. And words are not copyrighted. No one owns them because no one created them. So far, so facetious. 

But with a book cover design, this isn’t the case.

Traditionally, a book cover will be made of things other than words: so fonts, design elements, stock photography, or maybe stock illustrations. Someone else has created that stock or font set, and they wholly own the copyright to those things, and they give you rights to use them under certain terms with a licence. Those terms allow you to use them on a book cover, if you pay for the licence. Which is what a legit book cover designer does. What I do.

So does that mean that an author can now go away and copyright a book cover that I have designed for them? 

In a word: no.

Any designer saying otherwise is talking out of their posterior.

Because for the most part, you can’t exert the fact to a copyright office that you are the originator of the work. The photographer of the stock image and the font designs have done the heavy lifting here. With a copyright you need to prove that the work is unique and original enough (and there are no clear rules on this). And I would say with a book cover, the moment you tell them you are using stock images, they’re going to laugh out of the copyright office. Not original enough. End of.

Let me give you a real-world example.

Imagine you make a book cover with a stock image and you could copyright it. Then another designer or author comes along and finds the same stock image and pays for a licence and uses it on their cover. So your two book covers look similar in some way. What’s going to happen now? Are you going to say they broke your copyright, take them to court? They’ll just show the courts that you’ve both used the same stock image. Case closed. 

This idea of owning the copyright to a book cover is a popular misconception with self-published authors. A misconception that probably comes from the novel-made-of-words simplicity of being able to copyright a manuscript. A misconception, as a book cover designer, I need to correct on a regular basis.

But using licence material to make a cover gives you the Rights to use that cover on Amazon or wherever. This will become important in a second. There is a difference between Copyright and Rights.

So the question here is: if AI is creating brand new images that haven’t been seen anywhere else, why can’t I, or you as the designer, just copyright them? You created them right, they’re unique. 

At the moment the copyright office in the USA has already said ‘nope’. Something has to be created by a Human to qualify as an original work. There is no stipulation for AI generated work there. Test cases have been brought and failed. Probably setting some precedence. 

Here where the misunderstanding exists. Authors think they can copyright a book cover with stock images, and have heard you can’t copyright a book cover with AI images. 

I’m afraid you can’t do either.

So you’re in the same boat either way. To me there is no controversy. You’re out of luck both ways.

But if you’ve been following thus far, the question you would probably ask is this. Does the service that generates the AI artwork give me the Rights to use that image in a commercial capacity, i.e. to sell a book, in the same way as a royalty free stock image site. And the answer to that question is: yes! Or at least the service I use. I’ve read their terms, and it definitely does.

So this idea of owning the copyright to the artwork, is somewhat of a mirage for authors. 

“But I want to own my cover. I want. I want. I want.”

If you want to copyright a cover then get an artist to draw some image from scratch, or take a photograph for your cover, but to get professional work done you’re looking at hundreds, if not thousands of dollars. 

The reality is for cost effective book covers, you’re using either AI generated images or stock images, under the Rights, provided by either service, whether it be stock images or AI generated. 

What AI Definitely Isn’t: A Copy Machine

After reading a lot in the Self-publishing arena about the topic of AI Image Generation I see in some places there is a large misconception as to what it’s doing. 

It’s quite easy to say to yourself, “oh, it’s taking all the images it’s stored and just cobbling them together to make a new image,” as if it’s some sort of clever collage machine. 

That would obviously be stepping on someone’s copyright by doing that. It has to be doing that, right? I mean the images are too good to not be a copy of someone else’s work. A machine can’t make that.


But it’s such a great paradigm shift the immediate sensation is that it must be simply copying a human’s homework and. There just has to be some sort of plagiarism going on. 

No AI stores images that are copyright. It learns from them.

But how can it have learnt to draw that well? 

The above image shows you MidJourney’s improvement in 12 months. Yeah, it has learned to draw very well, very fast. Version five will hit at the start of 2023 and will be better and I think to myself how can it get better than version four!

But everything that it creates is completely from scratch from its learning. A good example to prove this point is to take an obtuse object and get it to design something that would have never been drawn in that style. 

Here are few examples:

This is the lithographic engraved style which was the only way to do images in books in the 18th and 19th centuries. Also, making music, I’m a bit of a synth buff, and this is no synth that I know of. The AI knows that synths have dials, keys and knobs. It’s deciding to make its own unique synth.

Likewise, I don’t think Xbox’s existed in turn-of-the-century Paris. Maybe less Absinthe would have been quaffed. 

I don’t think superheroes, let alone one called ‘Fish and Chip’ man were part of Picasso’s oeuvre.

All three examples sum up the reason why it would be somewhat impossible for the AI generation to just be copying images if it has to look in its database of all images that already exist. 

Yes, it can paint in the style of what it’s learnt. 

So how does it do that?

What AI Definitely Is: A Structure Analysing Machine then It Enjoys Lots of Arguments

We could get into the really nitty gritty of the technical ways of how AI learns and how it produces images. But some of that is even over my head.

But still, let me try to explain it in layman’s terms. 

It looks at millions of images, and analyses the tags of those images. And looks for repeated patterns. So it can learn the structure of things. Shapes, colours, line widths, composition, on and on.

As an example, if you give it enough apple pictures to look at it’ll know that apples are round and generally red or green. So records that structure. And it stores that ‘round’ and the colours in a database. Then you ask it to draw you an apple, it’ll draw something that’s round and that colour. 

Obviously, round and the colour is a simplistic way to look at it. The AI given enough images will understand other things about apples: stalks, where they’re usually found, that they’re not exactly round, that they can also be yellow or brown, what a half eaten apple looks like, etc. But you get the picture. So does the AI, given enough data.

The way the AI understands what you’re asking in it, and outputs good results from its ‘structure of things’ database, is achieved with something called a Diffusion Model. Basically you have two different AI’s arguing with each other backwards and forwards, until it has an answer for you in the form of a picture.

If you really want to get into the complexity of how this works you can read up about how diffusion models work you can read about things like Generative Adversarial Networks and Markov Chains. All subjects I got lost in because I’m a bit of a maths nut, and went down that rabbit hole. They explain what’s actually happening behind the magicians curtain. But yeah, rather full on, to say the least.

So in basic terms, the AI model has been trained to learn the structure of images. The things any human learns and knows. Cubism has lots of angles. Oil paintings have thick textured brush strokes. Rubens images are dark. A car has four wheels. The sky is blue. The grass is green. You see pumpkins on Halloween.

Literally anything you can think of that you would need to know to draw a picture, it’s been trained to know.

Controversy #2: How it Was Built (a Little Art History too)

So here’s where the controversy starts to hit. The image structure AI has been trained on millions, if not billions, of images from the internet, including a whole host of public domain and copyrighted works, from artists living and dead. 

When you hear the expression ‘copyrighted works’ most people start to feel, oh, there must be something illegal going on. The gut reaction is to remember things like Napster, BitTorrent, if you’re old enough. Something doesn’t feel right.

But as we’ve established here, what it’s doing is learning structure from those work. What’s the difference between you or I learning from art that has gone before, copyrighted or not? As an eight year old I remember learning to draw doodling X-wings and Tie fighters (very much youthful copyright infringing material). 

All artists and creatives learn from somewhere. When you first learnt to write, you might try to copy the style from your favourite author, until you find your own voice. Or as a comic book illustrator you might look at someone like Frank Miller and copy their style to improve your skills. It’s how we learn.

It’s also how the AI learns.

The problem here isn’t that a machine shouldn’t do what humans do, it’s the fact that the AI can do it on a vastly wider and faster scale. 

Let’s take an example. One that I guess is quite dear to me and one that I’ve been thinking about, when I’ve been considering the subject of AI. Picasso and cubism.

Stay with me, it helps explain what’s happening.

The first painting in this style was Les Demoiselles d’Avignon by Picasso, which he completed in Paris but it’s actually from sketches that he made whilst he was studying art in Barcelona. And it’s a close subject to me because I walk down the former-red-light-district-street of Carrer d’Avignon quite regularly (it’s about 1 km from where I live currently). Anyway, here’s the picture.

So we can say Picasso invented the style of cubism. It was a radical shift in modernism or short-lived trend, depending on your point of view. But it definitely was the latter. Because there were a good number of imitators that came along. 

In fact, in 2022, I was at the Sofia Reina Gallery in Madrid, which houses Guernica by Picasso — probably the most famous cubist painting. But that room was way too busy, like the Mona Lisa in the Louvre. So I mooched about and ended up spending time with this one instead.

It got me thinking about AI quite a bit. And it’s by another Catalan connected artist, it’s by a 22 year old Salvador Dali, yep that nutty chap. Hardly a style you associate with him. It’s like a young Dali is trying to be Picasso a good 15 years after The Young Women of d’Avigon. Catching the tail end of the cubism. See, I told you it was a trend.

So what’s this Spanish artist style-theft got to do with AI?

Well basically, we can probably draw a very clean line down through art history of styles being ripped off from one artist to the next, there is a clean lineage. 

As Picasso said himself:

As in, it’s better to take the best ideas from other artists and not just copy their work verbatim. Ironic side note: the original quote was stolen from TS Eliot, but I digress.

The point is that we’re at a point in technological sense where the AI can take everything that has come before it and learn ALL the styles. It can, and has, be trained to be every artist, by analysing all the work of all the styles. That’s just facts.

Think of the AI generation as the perfect art student that has been told to learn paint in every style that art history offers. It would take a human thousands, ten of thousands of years. And it cracked its task in less than six months. Or at least it’s 95% of the way there. Don’t get me started on how it does hands (at the time of writing).

And this seems unfair, very unfair. And quite a scary thought if you think about it for a length of time, as I have.

How Can the AI Get Away with Copying Styles?

In a word: you can’t copyright a style. 

This is where my little art history lesson comes in. Everyone copies from everyone else. Nothing is original, everything builds on what has come before it. The problem it seems to me is that AI has hyperturboed this development, and made it reality. 

As the old adage goes: we stand on the shoulders of giants. 

If you could copyright a style everything would have just stood still in art. 

So there is no copyright on style. It’s just the way it is.

So are we in a post-copyright world because of AI?

No. You can still infringe certain copyrights with image generation. Whether it be a famous person’s likeness (e.g. Messi, the football kicky man), or a character (like Charlie Brown), or product design, or logos, or a final photo (like a stock image). 

Put any of those things even on an AI generated book cover and you’ll still run into problems. But not the style they’re photographed in, drawn in, or what-have-you. 

If that’s the case, why all the controversy? Given that you can’t copyright a style and the AI just understands the structure of a style.

Controversy #3: Ethical Qualms

This is where it gets murky from an ethical point of view. And let me introduce you to someone I’d never heard of before I started using AI as a tool, he’s called Greg Rutkowski. 

We need to talk about Greg. He is a good example.

Greg is not a happy chappy. Greg is one of those artists that is alive to this day and is a jobbing artist. Greg’s paintings were on the internet. And these images were used to train the earlier AIs. So it then knew the structure of Greg’s style.

Then one AI user found Greg’s art and started using his name to make art because they liked his style. It made good fantasy art. So other people started using Greg’s name and his style became one of the most used prompts for a while.

He was getting Picasso’ed. People ripping off his original style. He wasn’t happy.

A few months ago he was probably one of the most vocal people out there shaking his fist at clouds. Maybe he thought he should be recompensed for the AI being trained on his images. 

I feel sorry for Greg, I do. 

But you didn’t see Picasso going down to Dali’s barrio and violently knocking on his door and asking for a fight or a big bag of deniros. Picasso was too busy moving onto the next thing.

If you ask me this is a complex issue. Here are my thoughts. 

I like to think of myself as an ethical chap. So I would love for the fact that living artists, who are used in prompts to generate art, got some recompense for their style being used. Seems somewhat fair to me. Will this happen? Who knows. I remain hopeful. Yet, quite recently it’s already going in another direction. More of that later.

Secondly, and more importantly, for me I’ve never used a single artist name to generate artwork, it just seems so icky. It doesn’t feel right. I’m a bit of a Golden Rule chap. How would I feel if I was Greg? So I’ve never used Greg’s name. Or any other single living artist for that matter when generating images for book covers. 

Plus as these things go: that sort of style became quickly overused.

One person uses Greg’s style to make a great image then everyone else just follows like sheep because they want to make great images too. And at one point there was a glut of Greg’s style (I know because I rate other people’s random images for free CPU time). 

Following the crowd has never been my style. Also if I have this fantastical tool why would I want to make obvious book covers with obvious styles. I don’t. And haven’t. It’s so much more powerful than just doing Greg pictures.

But here is where it gets even more complicated. 

Firstly, most of the people that are using Greg’s style, are they using it for commercial gain? Not necessarily. A lot of people use his fantasy style to make things for Dungeons and Dragons. Like character avatars. Should Greg get money when some dice-rolling 12-year-old is using his style for a bit of fun?

Secondly, does the AI actually make exact copies of Greg’s style when you use his name. Let’s do an experiment. Let’s take one of that original work image, shown earlier and try to replicate that using his name as a prompt:

To be honest with you, I actually don’t think it looks the same as his style and from an aesthetic point of view, I sort of prefer this AI image. It has more of a vibe to it. Sorry, Greg. The AI has already surpassed you, even if I use your name (which I would never usually do). 

Thirdly, as I mentioned you can’t copyright your style. You’ll see from my example, is it really that close to Greg’s style, anyway? Some elements of structure might remain, like the shapes of a dragon. And I have a sneaking suspicion that Greg didn’t invent the shape of dragons, let alone copyright it, otherwise he’d be sitting pretty on all that Game of Thrones royalty money. 

And at the time of writing this article it’s actually really hard to replicate his style. This might be something to do with the Spawning service (we’ll get onto that soon). 

And finally, this is where it gets interesting. Remixing. 

As we’ve established, down the annals of history, arts have borrowed or stolen from multiple other artists and developed their own unique style. Dali himself, as well as his foray into cubism, started knocking about with the Dadaist, and the Surrealist. All these inputs made Dali what you remember him to be: that melty clock bloke. 

This is what AI is really good at. Here are some examples of the same image, where I’ve added a second artist. So a cocktail of Greg and another artist.

All of them become their own style. Nothing at all like Greg’s style. Or any of those other artist names I’ve used. It’s become its own thing.

It’s almost like, with AI, styles have become ingredients in a recipe. It’s a way of thinking that never existed before AI image generation. 

But do I feel for Greg? Yes I do. Imagine spending 10-20 years of your life developing your own voice, and taking hours to paint a picture, and then AI comes along, learns your ingredient in months, improves on it, and then can be thrown into a recipe. 

But that’s just where we’re at. There is no putting the genie back in the bottle.

It’s like the moment when all those car factory plant workers were laid off when robots came along. There isn’t a point where Ford or GM suddenly said yeah, this is awful for the workers, let’s get them back in. Those jobs just don’t exist any more. And civilization has moved on from that. It’s harsh. But that’s just what happens.

We’re in that radical technological shift moment in time.

Let me Introduce You to Spawning

And here’s the rub. I think I might have been lying to you.

I’m not entirely sure the newest version of the AI that I use even has Greg’s art even in it. I think it’s more of a generic red dragon.

Because things have already changed pretty fast to deal with this ethical issue.

So when some living artists kicked up a fuss, a service came along, called Spawning, advocating for artists, and created a database of those that wanted to opt out of the data that was used to train these AIs. 

And every time they update any of these AI Image Generating services it needs to be retrained from scratch. And as they’ve gotten more proficient at training the AIs (as shown right at the start of this article), they need to rely less on living artists’ work. 

So I suspect Greg’s style is on this block list now.

So everything has already quickly caught up with this annoyed artist question, ethical concern. But it doesn’t stop media outlets churning out contentious content to grab eyeballs. Controversy sells. Proper journalism is harder to do.

It sort of makes me sad. I would have preferred for Greg to get a bit of a kickback for his original training of the earlier AIs, even if I never used his style.

And now we’re going to stop talking about him.

I don’t even like his style, far too flowery for my tastes, a bit too Monet. Was Greg ever inspired by Claude Monet? Now that would be ironic.

Controversy #4: Signatures & Watermarks

Before we get past all the controversy, here’s another interesting aside to this ‘trained on artist copyrighted works’ which is rather messy, to say the least. 

Greg and other artists, right down the ages, have typically relied on a rather old school way of protecting their works. They traditionally signed their works, or in a more modern way, watermarked their works.

So knowing what we know in the way AI learns the structure of images. What do you think happens when you ask the AI to draw a picture from such an artist? The AI has learnt that that part of the composition is part of that artist’s structure and replicates it. Because it’s daft.

Me, I’ve never had signatures pop up in my AI creations because I generally shy away from living artists. But everyone out there in copying internet land is using artist prompts where this has been learnt.

But this being some sort of proof that it’s cobbling together images from copyrighted work are just that: cobblers. It’s the AI simply trying to sign its own paintings, because that’s the structure it’s learnt.

And that’s it with the controversies. 

I much prefer to talk about myself!

Where I’m At

As a cover designer, AI has been a godsend. I’ve loved playing with it. It’s incredibly useful as a tool. And I use the word ‘tool’ judiciously. Nothing replaces understanding, experience, imagination and hard work. 

It’s easy to think of AI image generation as a magical thing, you just simply type something in there and it gives you the most amazing images that are perfect. If anything, to get good work out of it, it’s been a very steep learning curve for me. Getting what you want takes time to perfect the descriptions, nudge the argument between natural language and the structural nature of images this way and that. 

Here are a few facts about me and AI.

Firstly, I’ve generated about 15,000 images and probably found about 1,000 of them actually useful. 

Secondly, when I’ve used it for a standard commission with my clients I probably spend about twice as much time to generate the right image to use than it would have taken to find suitable stock images. But the results are definitely superior for certain genres of book when it comes to AI. And this is another point. Finding the right way to prompt it for each genre is a whole new set of skills I needed to learn. 

But what it has been really good for is opening up my own brain and thinking in a brand new way about my own semiotic connections, expanding those. I’m sure my design brain is a little bigger after these six months.

But there is much more that’s great about it other than the expanded number of synapses in my stupid fat head.

What I Love About A.I. Generation for Book Covers

But for me this is where it gets interesting, for what AI Image Generation can actually achieve over stock images. There are loads of great advantages to using it to make book covers. 

It’s a tool that goes beyond what humans can do if you open your imagination wider than what’s come before.

Uniqueness. People do want their book to turn up in a totally one-off design. AI Image Generation is really good at this. In fact, the fact it makes new images every time sometimes gives me too much choice! But allows authors to rest assured in the fact that someone won’t come along with the same stock image used on their cover.

Closer to the Author’s Ideas. When I’ve worked on commissions using it, I’ve been able to get closer to what authors want. “Yeah, my main character is a mixed-race, twelve-year-old angel in a magical dress made of crystals.” No problem. I can’t imagine the impossibility of achieving that with stock images. It’s now easier to match an author’s wonderful imagination. 

I Can Give Authors the Original Images to Play With. In the past, when I’ve used stock images to make covers, I can provide a client with the final work, but not the stock image originally used, simply because that goes against the standard licensing terms. With AI generated images that’s now possible, legally. Which I’m pretty cool with, because I’m cool Fonzi. Although if you want me to make other bits and pieces of advertising to go with your cover, I can do that too, from my Design Extras page. You know, properly designed.

Representation. For me this has been a bugbear down the years. The limitations and biases of the originators of stock images always became my limitations to provide for authors. Me, I’m a ‘Bill and Teds’ person: be excellent to one another. I’m happy to work for any race, creed, able’ness, gender, sexuality and kink, as long as you’re being excellent too. AI opens up a lot of possibilities and this makes me really happy that I can provide for everyone. Whatever you want: female pirate captains, spies in wheelchairs, black spacecraft captains, bored purple aliens working in Walmart, Hasidic Kung Fu, or whatever crazy things you can imagine to write about! The AI doesn’t judge. Neither do I. 

AI can easily work out and draw things that are too hard for Humans. You would have thought that anything that can be thought of in an artist’s mind can be visualised and drawn. But this is simply not true. Think of something like a fireball firing across a lake, whilst the water is being disturbed by an iridescent green glowing dragon that breaks the surface. Getting all those reflections, diffractions, and light sources would be virtually impossible. Or any other complex mathematical concepts. AI takes a lot of this stuff in its stride. To me, it’s the most interesting area of aesthetics when it comes to AI image generation. When I see one of these images the hairs on the back of my neck stand up. 

Colour Palettes. One of my absolute favourite things about design is simply: colour. I love playing with colours. Having things that really catch the eye for covers. Something that stands out from the crowd. AI is stunning at this, being able to take an image that might be a common trope and hit it with an interesting palette.

The Madness. I have way more fun with MidJourney just giving it all manner of utter crazy prompting words. I try to confuse it. For some reason this works particularly well. Emojis. Odd character from the full character set like these: ¤ ʨ ₱ ╦ ◙. Or just make up some ideas that seem to make sense linguistically: MinatureCore, MeltPunk, CandySplat, Swizzollo. It all seems to give interesting results.

Where We’re All Going

There is an interesting point I want to make here in closing, about where we’re heading in the future. I’m actually a rather philosophical sort of chap (in the classic sense), so I’ve spent a lot of time with my thoughts on this subject. 

The way I like to think about it is this.

When we talk about AI there are two different concepts at play here in a sense, what I like to call ‘creativity’ and ‘artistic impulse’. 

Creativity says, go make something. I want the final thing. The joy is making it. I spend hours writing a book, writing a song, painting a picture, cooking a meal. And the end result is what you’ve ‘created’. The object. Which is a fine pursuit in life. I do a lot of it.

But there is a deeper way that people make things, which I would call artistic impulse. You have an imaginative idea or deeply felt feeling or brand new thought you want to express, it comes way before the creative part. In fact, the creative part of the process is simply to fulfil that artistic impulse. And generally the end result will not be what that artistic impulse was at the start of the process. It will change and you will be changed because of that journey. I do a lot of this as well.

What AI will never be good at is: artistic impulse. Ever. There always has to be a seed, that’s human and imaginative. 

What AI will replace is that creative slog of the making, eventually and with everything. It’s all in the post, not just images. Think 3D worlds, typefaces, music, fashion, architecture, product design, recipes, and yes, writing.

It just got very smart very quickly. Very good at processing.

If you ask me how far off we are from:

“Write me an 80,000 word book in the cut up style of William Burroughs, about AI, set in Barcelona, with a subplot surrounding a stolen Picasso painting and second subplot involving the fantasy artist Greg Rutkowski, in the cosy mystery genre with techno thriller elements.”

I would say, at most, three years. It won’t be perfect, might need editing but yeah, but a perfect manuscript might be another three years off. But someone still needs to come up with that crazy idea. The artistic impulse.

If you had to pin me down I would say, you’ll probably have some sort of AI book cover design service in about 3-4 years. Will that mean I’m out of job? I think not. Because someone will still need to come up with what actually makes a good book cover, and that’s a skill on its own. An artistic impulse to set the creative wheels in motion.

Or maybe not. Who knows.

Things are changing fast. And I’m trying to keep ahead of it all. But at the same time I think it’s good to share my thoughts as well, so I hope that this has been somewhat of an education, clarifies some points for you all in a simple way.

Before I go, it’s worth pointing out that this article was written by a human being …

… maybe. 

No. Joke. It definitely was a human, that was me, 


PS I swear on my life the next blog post will be shorter. If you’ve made it to the end, well done you. Go reward yourself! Maybe a chocolate bar or something.

  • Recent Posts

  • Archives

  • Categories