Pragmatic Hot Takes: The Industrial Revolution and AI Art
3 years ago
There's a few hundred years between these points in time, so bear with me as we go for the gold in the metaphors and logic leaps categories, with a side event of unfairly boiling down incredibly broad and complicated historical subjects into a semi-digestible pseudo-intellectual sludge.
Back in ye olde days, there was this little thing called the Industrial Revolution. Suddenly, one man, woman or (infamously) child could do the work of ten serfs. The good: such a massive leap in productivity made goods far more affordable for the common man and helped kickstart the middle class! The bad: the death knell of the Cottage Industry, awful working conditions, massive pollution generation, kickstarting the middle management class... Look, you'll just have to trust me that it was generally a good thing for society since otherwise I wouldn't be delivering this to you over the 'net.
Anyways, at the time people weren't really a fan of the whole "Your livelihood has been replaced by a machine that can massively outproduce you". There was a rash of incidents in France (if you ever need to learn how to protest, read up how the Francophones go about the matter) where they threw their wooden shoes into the delicate gears and cogs that made the machine of industry turn. Fun fact, those shoes were called Sabots, thus birthing the term sabotage.
I imagine you're starting to piece together the comparison between then and now. What once was the creative output of thousands of families creating bespoke clothing and other forms of artisanal crafting became mass production and commoditized, a flood of cheap goods that priced out the humble seamstress and tailor. "We've seen it before, they're coming for us now!" artists yell, and they're not entirely wrong.
There are going to be many knock-on effects; many artists in the middle will find themselves overshadowed by the fidelity of AI art, and it's likely those who would have trained and practiced to become great themselves will no longer feel the need to invest the time and energy when they can make something good enough with some keystrokes. Similarly, commissions--lifeblood of the chronically funds-ailing artist--will dwindle, as those who once bought them will move on to cheap or free services that can pump out a dozen pictures in a fraction of the time that satisfy their immediate craving.
Here we hit the nadir of the doom-saying: Artists will be pushed out, commissions will slow, there will be accusations of using AI art as a crutch for artists trying to pass it off as their own work (this has already happened). It will be a monumental upheaval of the artist space, people will fight bitterly about it, there will be an internet flamewar.
And it'll all be worthless because the war was lost before they even realized it was being fought. There is no putting the genie back in the bottle, just as there wasn't when sewing machines crushed the loom. It's not just artists, too, but many creative professions; the advent of the AI generated voices was celebrated when it was used for silly YouTube TF2 videos or animations but now we're seeing professional voice actors, like James Earl Jones of Darth Vader fame, selling the rights to their cadence and tone as well. Writers (like myself), music composers, and undoubtedly many other professions will be rendered completely and utterly irrelevant by the advent of AI!
Except we won't. After all, we still have seamstresses; my late Aunt was one, in fact. I got a suit for a party from a tailor just last month. These professions haven't disappeared, and neither will the likes of artists or other creative pursuit professions. An AI, for a long while yet, will not be able to perfectly capture all those little intricacies that a commissioner wants for their OC, or the perfect pose, or the right colour tones, or form consistent picture series for a comic. If anything, I imagine there'll be a burgeoning business of hiring artists to edit the products of AI-generated art to make them just-so. The ability to create art with them will instead be judged by the creativity of the individual generating the prompts, where even someone with physical or mental disabilities can conceive of wonderous landscapes and have the machine help them create a work of beauty. Perhaps I'm biased in this regard myself; I have coordination difficulties that make creating art of my own difficult, ergo being a writer instead, but I digress.
What we are witnessing is the democratization of fine arts, making them attainable to the common individual. Writers able to source pictures for the stories, artists able to generate writing for their work, so many new avenues opening. It's going to hurt, but growing new wings isn't a painless process. I've already seen posts of people cutting their nose to spite their face when it comes to AI art but the simple fact of the matter is it's here to stay, so gnashing our teeth and cursing the heavens is only going to give us a sore jaw and raw throat. Learn how to use it, how to exploit it. There'll still be a market for skilled artisans for many years, generations even, yet; capitalize on how it can help you and ride that wave.
Plus, we don't have wooden shoes, and they don't have gears and cogs to throw them into, which complicates stopping the process even if we were to make the effort.
Back in ye olde days, there was this little thing called the Industrial Revolution. Suddenly, one man, woman or (infamously) child could do the work of ten serfs. The good: such a massive leap in productivity made goods far more affordable for the common man and helped kickstart the middle class! The bad: the death knell of the Cottage Industry, awful working conditions, massive pollution generation, kickstarting the middle management class... Look, you'll just have to trust me that it was generally a good thing for society since otherwise I wouldn't be delivering this to you over the 'net.
Anyways, at the time people weren't really a fan of the whole "Your livelihood has been replaced by a machine that can massively outproduce you". There was a rash of incidents in France (if you ever need to learn how to protest, read up how the Francophones go about the matter) where they threw their wooden shoes into the delicate gears and cogs that made the machine of industry turn. Fun fact, those shoes were called Sabots, thus birthing the term sabotage.
I imagine you're starting to piece together the comparison between then and now. What once was the creative output of thousands of families creating bespoke clothing and other forms of artisanal crafting became mass production and commoditized, a flood of cheap goods that priced out the humble seamstress and tailor. "We've seen it before, they're coming for us now!" artists yell, and they're not entirely wrong.
There are going to be many knock-on effects; many artists in the middle will find themselves overshadowed by the fidelity of AI art, and it's likely those who would have trained and practiced to become great themselves will no longer feel the need to invest the time and energy when they can make something good enough with some keystrokes. Similarly, commissions--lifeblood of the chronically funds-ailing artist--will dwindle, as those who once bought them will move on to cheap or free services that can pump out a dozen pictures in a fraction of the time that satisfy their immediate craving.
Here we hit the nadir of the doom-saying: Artists will be pushed out, commissions will slow, there will be accusations of using AI art as a crutch for artists trying to pass it off as their own work (this has already happened). It will be a monumental upheaval of the artist space, people will fight bitterly about it, there will be an internet flamewar.
And it'll all be worthless because the war was lost before they even realized it was being fought. There is no putting the genie back in the bottle, just as there wasn't when sewing machines crushed the loom. It's not just artists, too, but many creative professions; the advent of the AI generated voices was celebrated when it was used for silly YouTube TF2 videos or animations but now we're seeing professional voice actors, like James Earl Jones of Darth Vader fame, selling the rights to their cadence and tone as well. Writers (like myself), music composers, and undoubtedly many other professions will be rendered completely and utterly irrelevant by the advent of AI!
Except we won't. After all, we still have seamstresses; my late Aunt was one, in fact. I got a suit for a party from a tailor just last month. These professions haven't disappeared, and neither will the likes of artists or other creative pursuit professions. An AI, for a long while yet, will not be able to perfectly capture all those little intricacies that a commissioner wants for their OC, or the perfect pose, or the right colour tones, or form consistent picture series for a comic. If anything, I imagine there'll be a burgeoning business of hiring artists to edit the products of AI-generated art to make them just-so. The ability to create art with them will instead be judged by the creativity of the individual generating the prompts, where even someone with physical or mental disabilities can conceive of wonderous landscapes and have the machine help them create a work of beauty. Perhaps I'm biased in this regard myself; I have coordination difficulties that make creating art of my own difficult, ergo being a writer instead, but I digress.
What we are witnessing is the democratization of fine arts, making them attainable to the common individual. Writers able to source pictures for the stories, artists able to generate writing for their work, so many new avenues opening. It's going to hurt, but growing new wings isn't a painless process. I've already seen posts of people cutting their nose to spite their face when it comes to AI art but the simple fact of the matter is it's here to stay, so gnashing our teeth and cursing the heavens is only going to give us a sore jaw and raw throat. Learn how to use it, how to exploit it. There'll still be a market for skilled artisans for many years, generations even, yet; capitalize on how it can help you and ride that wave.
Plus, we don't have wooden shoes, and they don't have gears and cogs to throw them into, which complicates stopping the process even if we were to make the effort.
"No, Artists are not being "replaced"; there will always be a place for the truly talented artist... There will probably be fewer mediocre ones though. Because well... if your art CAN be reasonably replaced by a machine, you might have something to worry about. If your skills can bay the bills; then great, you'll still have lots of work. If not... The fast food service industry awaits."
Having lived long enough to remember the world before the widespread adoption of the internet by the average person; I can quite vividly recall people swaying "computer art isn't real art!" when paint programs and tablets first started showing up. But well: look where we are now. For those who can adapt; these algorithms are going to be awesome tools.
What WILL probably go teh way of the dodo though are things like "storyboard artists" and that sort of initialc oncept work; or rather, you'll have ONE person and their pet VI doing it instead of five.
I think storyboard artists won't go away; their work is sketchy and is thus quick to produce regardless, and depends on the context of the previous panel to make the next, something AI art is going to struggle with for some time yet. Concept artists, though, those could certainly take a hit.
You're absolutely right in that it will have a significant effect that cannot be stopped, and that parts of the profession will take a hit but as a whole will not die out. Even if those roles were in danger artistic roles that cover creative and management positions will need to be humans, and people with art knowledge could move to those jobs. The technology has limitations even if most of the current results are more of a proof-of-concept, and it's kind of ironically motivating to see some online discussions prove the difference a critical eye has on reception.
The thing I'm personally interested with the most is the legal angle of AI art. It's pretty clear that in terms of hobbyists and those who just want their idea on the screen that the existing applications are good enough for their purposes. Have the AI make an image, then either use it, touch it up, or use it as a reference for your own pieces, couldn't be simpler. The problem I see is when it starts hitting the professional space, who owns the art? Is it simply the company that created and trained the AI, making a service for individuals and agreements for large professional places necessary to resell? Alternatively, would the buyer's copyright apply to prompts in some way since the artist is a program, meaning neither human party could legally use it without mutual terms and conditions or open-sourcing? There's even a small argument to be made that the AI itself owns the art, though I don't think the possibility of robot rights is at the forefront of peoples' minds. Regardless, at this point it's rather inaccurate for one to point to an image they gave a prompt for and say they alone made it.
There's also the real possibility that artists in some way prove that the images generated by companies that would sell trained models violate licensing or trademark, since even though the common AIs are trained through diffusion and output more fundamental patterns, they use datasets from web crawling. Notably for this possibility, Getty was able to force direct image links in Google image searches back in 2016, so stock photo organizations may have some significant sway in commercial use of AI. Obviously, this doesn't hinder AI development as a whole, but would mean that it would be more in the hands of developers to create their own image database to pull from. That potential, in addition to that this method seems to be more effective at reproducing based off an individual art style or character brings the possibility that the future of AI art may be more in the hands of artists than we currently think.
My personal guess is that the common use-case for this will be more of a tool for all types of artists, rather than a replacement. People who make a living off art will lose customers to the programs, but in turn will be able to use them to speed up the processes they use and be proactive, just as you say. Ultimately, having knowledge in art and the field as a whole will still be highly valued, though I'm sure attempted data poisoning will happen anyway.
Well, here at the potential end of FA as a viable platform for my fetish de jour, I figured I 'ought to finally give you the response you deservioli.
One important note that I didn't know of at the time of writing is that the deep learning algorithm doesn't examine a bunch of pieces and then fabricate it like a human, but rather overlays them and then dynamically filters them together like someone mixing paints. It's an important distinction, as it means if you removed the reference material it couldn't produce anything; it hasn't learned to draw, it's been taught how to cut and paste on a pixel-by-pixel level, in effect. Obviously someone who knows more about it would scoff at my surface-skipping analogy, but it does the job and highlights a major factor in how much it can be considered 'art', I feel.
It's still an incredible tool, but I feel any works fed into an output need to be properly credited and AI-generated art only used as reference material, not an end product, as a result.
Thank you for your detailed and thought-out reply! It deserved a response from me long ago X)
The response is certainly appreciated, and it's unfortunate that the site's staff forces spurring this on with the increasingly vague and incontestable rulings. I don't know specifically where you may be moving forward to, though I do hope to bump into you wherever that may be.
I agree, it's an important distinction that while the common explanation is that it's learning like a human with neurons and connections, it's much more complex and different than that while being difficult to explain in a meaningful way. Your explanation of cut-and-paste on a pixel by pixel level is a pretty good overview of its capabilities and design principles, and though tying in the fuzzy logic portion of some of these systems might be difficult for that particular analogy, I think it's quite great for getting your point across.
The ethics questions of scraping for training data and not crediting when the program is somewhat tightly tied to said data is certainly one of the more pressing concerns for the technology. I can only hope that in the current time it is either used as a tool for reference as you describe or people are generally truthful with when a result is a generated work, touched up or otherwise. With its evolution, hopefully the developers will create new methods to credit the works used or make it less reliant on what it was trained on rather than developing better ways to obfuscate influence.
I'm happy to get the response now! It's been a wild ride with this complex technology so far, and I am quite glad to discuss with you and to hear that my comment caused thought.
Welcome, and yeah. But when has FA ever been run competently? At least this is helping push people to better websites (hopefully).
Heh, thanks. It's a bit mired in Layman's Terms, but it gets the idea across well enough.
I feel like the output should forcibly include the URL the art pieces were sourced from so it can at least be traced to its original owner.
It certainly has been, and I'm excited to see where it goes, although it clearly needs some better controls in place for ethical use.
It would be much better if the models showed what is used on output, though matters of scale do make it difficult to do things like encrypt the credits in the image. Even just an open database of images used and the program's watermark on the output tracing back to what images went into the black box would be nice in comparison to the current situation.
The trajectory of new technology is always fascinating and exciting to watch, and even more so to experience.
Just gotta wait and see what developments come if it machine learning!