**New post about my revised stance on AI**
The following post does not reflect my current views.
-----------
If you would have asked me that question 2 years ago, I would most likely have said:"Not in this century".
Speaking of digital art of course.
Now that I'm in a beta-group of stable diffusion, I'd say it might be a month or so away from happening.
First off; I don't want to scare anyone and I'm personally excited about the development. But fact is; the release of a public-domain text-to-image AI that is capable of creating artworks like the following, is going to disrupt many lives:
 |
Image top -right @vinyvince / landscape image @Nas |
I may refer only to Stable Diffusion in this post as all images shown here are from SD (except the 2 hand painted ones). You can read more about the Tool in question here: https://stability.ai
You can have a look into this Sub from reddit to see what other users come up with: https://www.reddit.com/r/StableDiffusion/
As far as Midjourney goes, I'd say no, but that is due to its extreme limitations which I learned about just a few weeks ago when I joined the 2nd wave of beta testers for Stable Diffusion. What I saw there is what AI is capable of and it is scary in some way but in the right hands it can also lead to amazing results and liberation. Seeing that made me realize HOW limited Midjourney really is, because 90% of prompts that looked amazing in Stable Diffusion were even with hours of re-prompting not possible to recreate with MJ.
Dall-e 2 goes in the same direction, the censoring and limitations do more harm than help.
The founders of Stability.ai speak about democratizing but I'm not exactly sure what they mean by this. Academics and regular people often speak a very different language. In my opinion they want to make it freely available so that more diverse art and products can be created. Because as it is now with NightCafe and Midjourney, they all have specific looks like traditional media. That is intended to have some kind of a trademark. Like watercolors have a very distinctive look you can recognize a Midjourney image from miles away.
Personally I really look forward to Stable Diffusion as it could change my workflow and allow for many different styles of art. This could open up new markets too! One new way for artists could be to generate artistic styles that can then be licensed to companies for games, products, whatever.
I also see a chance to be able to think about doing commissioned work again. A thing that I had less and less time for in the past few years till a point where I only have 2 companies left I do work for and I don't take on new ones atm. Stable Diffusion trained with my personal dataset could change that very well.
One reason why I rejected so many commissions / opportunities was that I could just deliver 50-60% opposed to what I could do when working on my own projects. With A.I. I imagine being able to deliver at least 80% of my best work and I could have fun doing so. Because believe me, delivering only mediocre work that takes 40 hours and more is no fun at all. That is business.
But is it Art?
Then there is the ethical argument of artists or non-artists, complaining that this is not Art anymore when just using a prompt to generate images.
My question here is: "What if in 100 years no one could imagine to do art with hands because that could be too difficult, IDK. Imagine then they would not even do text-input-to image-prompting but instead they would just imagine images directly because our brains are connected to one big machine... Would that be art by our limited standards and understanding?
Is art only art when you are educated?
Is art only great when you have put at least 40 hours of work into a piece?
Is photography art? (Because you just click on a button)
These are important question that I ask myself. For me the process is not important. In the early days of deviantArt, I had 10-minute sketches that got 100 comments and works that took 50 to 60 hours that no one bothered to look at.
From a copyright perspective, there is authorship and ownership. In that regard there is not much difference between writing an Haiku and owning it and writing a prompt to a machine which generates an image out of your text.
Some artists are concerned about the copyright issue because of the similar looks being produced if the machine is prompted to work in the style of one particular artist. And that is interesting for non artists as fans of that particular artist or model or photographer. Artists on the other hands would always chase to combine 2 or more artists together and see what that does - and that is a creative process and the result is something completely new.
In addition - combining just 2 artists in a prompt allows for an infinite number of possible creations. Navigating through this AI-journey is a like navigating through a jungle of unknown territory.
In terms of copyright you have to know exactly which is which in order to prevent accidental damage and liabilities. Because in the end - if you are getting sued because of violating someones rights through the use of AI, you can't blame the machine for it and that solves the question about authorship once more.
The ethical question lies in the way we as artist perceive the process. We need to get used to this new way and decide for us if this is still art or is it cheating?
Because the viewer (not in the know) might be moved or really like an output, does confronting the viewer afterwards lower their perception of that initial value to them? I don't think so.
Whenever I told people on a convention that I did not painted the landscapes on the Ancient Kaiju Project, but rather painted over it in the style of the existing work a piece of popular culture, they were even ore intrigued by the story and the process. Not less.
Because the viewer (not in the know) might be moved or really like an output, does confronting the viewer afterwards lower their perception of that initial value to them? I don't think so.
Whenever I told people on a convention that I did not painted the landscapes on the Ancient Kaiju Project, but rather painted over it in the style of the existing work a piece of popular culture, they were even ore intrigued by the story and the process. Not less.
At the moment it's artists against artists which make me sad and I understand that some people are afraid but fear is never a great way to build a future, it leads only to black and white thinking, like the discussion here on a recent dA posting shows. It really leads to nothing.
Well, before A.I. there was tracing. Or a photograph was considered cheating around 1900. Photobashing in Photoshop was considered cheating a few years ago and Kitbashing in 3D was also considered cheating.
The point is: it becomes cheating if you are in a competition for that certain category that requires participants to draw by hand or create from scratch.
However, we are not in that competition as a whole. In the entertainment industry, time is money and if you can churn out 10 images an hour thats good, if you can churn out 100, it is even better.
In a market where you want to sell prints, the buyer decides what they like and what not depending on many factors. And one is not if you spend at least 50 hours on a painting.
One positive thing about this rising quality of A.I. is that it becomes more likely that rip-off companies and the likes rather create their own work based on A.I. instead of ripping the artists. This could have very different consequences; of course it is better for well known artists because of less losses and less dmca's but also less chances for rising artists because like it or not, a plagiarism is also a promotion of that particular work which in return creates a demand. Less demand = less inquiries for that particular artist = less popularity. This all is part of the democratization that is upon us. It will be hard to foresee how it exactly plays out but that is what I learned over the years.
Will there be jobs at Risk?
Sure like with every wave of new technology or trends that come up and go. We are at a point where striving artists at the verge of starting a business versus keeping art as a hobby - have to think twice. Either you have the balls to start a business or you go the safe route. It is the same movie with the red-pill and the blue-pill, which one do you take? I understand the concerns but starting a freelance business is not much of a difference today than it was 2009 when I started.
But that is nothing new, the advent of this technology just changes the rules yet we have to wait how the dices turn out because they just began rolling.
And I'm sorry to say that, but you can't opt out of it. As digital artist you have to be very well known and better be very good (and fast) in order to stay away from A.I. Everyone else who has to make a living as artist better get their hands on it very early to stay in the game.
You will sell less art when everyone is an artist!
I disagree. Competition is good for business. In crowded NFT-Markets with so many trust-related-issues, that is a problem. Selling prints on a POS like a convention, not so much. Also; even if you can cook yourself like a chef, sometimes you just like to eat out and enjoy that.
So not much would change in that regard - or only if you are used to sell an a3 print for 60 €, then you might get used to just charge 15 -20 €.
Traditional artists are safe, no?
For a certain time and for some areas maybe. But also traditional or analog working artists usually sell prints of their work. That market will change for sure with billions of new and amazing works flooding the market within the next years.
In the end, my aim is to create the best images I can. For this I learned to work with Zbrush, Painter, procreate and Photoshop. Before that I learned watercolors, colored pencils, charcoal oil, acrylics and airbrushing. Always striving for the next best painting because the next one is always going to be better (for an artist;).
In my ancient kkaiju project, but also the handpainted inceptionism, I tried to digitally learn by painting subjects, the original artists would never do, like the following pieces from Brom, the second one is an interpretation of Widowmaker from Overwatch painted by Brom:
There is more of these if you like to see: https://fantasio.work/albums/65214
So for me that style-transfer or inception is nothing new I just did that by hand and I know it is really hard work not really appreciated by most. I did not expect that Machine-Learning is ready to do exactly that in this pace. In the following pieces, I did just what I did in the album above but with the help of Stable Diffusion. The Concept for the prompting was "figure xyz as Overwatch character painted by Greg Manchess and Gil Elvgren":
As a self proclaimed "Expert" in the field of hand painted style-transfer, I'm officially stunned!
In my book, these are all amazing concept sketches from which I could work from. Not perfect if you look closer but they are actually OK as they are. Many concepts I have seen in Art Books are way below this when it comes to accuracy or anatomy. Right now I'm using these unedited portrait sketches as content for my social media to raise awareness for the coming storm.
I hope to get as many reactions and feedback as possible to add in a follow-up blog post.
For now I'm hopeful that this kind of technology will create more opportunities than it destroys. It is a sword with more than 3 edges so it will be misused and that is a topic for a complete different article.
Please let me know how you feel about this coming change in a comment.
Hi Oliver,
ReplyDeleteThis absolutely is a game changer. This will allow me (and Griffon Lore Games) to publish the amount of art we want (we only put in 1/4 of the art we need due to costs). This in turn will allow us to spend more money, or at least targeted money, on commissions. It's a win-win.
Hi Anthony!
ReplyDeleteI agree! Thanks for the comment, appreciated! It is what I imagined it might be very useful for. A friend / illustrator having to do 100 illustrations for a roleplaying book at $30 per piece. Working on that traditionally for 2 months if he is fast is just not a sustainable business even for freelancing.
Reducing that time to 1/3 would be a win on both sides in my opinion. There is a lot potential without harming the industry or artists but it has to be played fair.