Trovate qui la versione in italiano di questo articolo.

#Midjourney, #StableDiffusion, #NovelAI, #OpenAI… you name it. There’s a new one coming out every week.

Picture-generating algorithms are here, and they are putting at risk the entire sector of illustration and concept design for creative professionals. Social networks are overflowing with tons computer-generated pictures, and an equal amount of artists rightfully complaining, ranting or downright heartbroken about it. I should know: my wife is one of them. And I myself feel kinda worried about it.

Now, this isn’t going to be a technical article – I have not the competence, nor the actual experience with AIs to talk about this phenomenon in such terms. I would rather try to analyze the potential impact of this technology on a type of business I’ve worked myself in, and rule out a few proposals of my own for potential approaches on how professionals might deal with SkyNet-with-a-paintbrush.

What is AI and why it outcompetes you as an artist

Ok, so, picture this: you’re an artist and you love making art. Maybe you consider yourself a good, accomplished artist. You’ve started drawing at a very young age, you’ve chased your dream by going through art school and you’ve spent ten, twenty or more years of your adult life studying, working hard and taking crappy jobs with impossible deadlines to perfect your skills. You’ve gone through the hassle of contacting publishers, agents and building your personal brand on social networks and conventions, all to become a recognized professional; or maybe you’re still working on it because, what artist feels like a completely accomplished professional who has nothing left to achieve or learn?

One day you sit at your working desk, you open the Internet and read an article that says that an AI will make you lose your job for sure within a few years. Why? Apparently, because someone invented a software that isn’t a tool for a drawing artist to use, but that makes a computer draw in your place, which means you are a commodity now for many companies who want to cut down their expenses. A PC draws better than you (well, until it doesn’t), a thousand times faster than you, for a thousandth of your usual fee. Also, it doesn’t complain about retakes, it doesn’t get ill, it doesn’t have kids who get ill, it doesn’t need holidays, it doesn’t divorce, and having many clients doesn’t make it skip deadlines. So, you suck in terms of being competitive. Companies should know better than hire you: they should totally hire Mr. AI.

No matter how you put it, this bites. The very idea of being replaced by a heartless machine is painful to any artist, it is diminishing of all the hard work you’ve done, and it doesn’t feel fair at all.

Even more so, if you think that the AI may very well have sampled the art you’ve put online, analyzed it bit by bit, and it’s reshaping it at the push of a button to create a new image – a new image that has YOUR bits in it, but for which you won’t get paid. Because that’s what AIs do: they harvest images from the web, sample them and use the data they collect to produce new combinations – fully artificial images that look new, but are, in fact, very clever remixes of existing artwork. WITHOUT the original creators’ consent. And without adding customers to their brand, or paying them royalties.

I don’t need to tell you how damaging this current approach to AI technology can be. Companies are already creating book covers and concept art via algorithms, and artists who depend on their art for a living are subsequently losing commissions. Also, this sends the message that artists are a replaceable part of the creative process. Now think of replacing your doctor, or your favorite politician with an AI; would it still feel like playing around with a computer? How can we artists deal with such a big loss in our ability to compete? And, is AI really bad or it can be used for the good of artists, too?

AI, friend or foe? Well, that depends.

Now, I understand I may have been overdramatic with my opening so let’s tone this down a bit and face the issue as rationally as possible.

Many big innovations sometimes start as child’s play in some nerd’s garage. Apple. Windows. Napster. Human ingenuity is limitless and it has given us the most amazing technical tools which have profoundly transformed our workflows over the last decades – just think of the transition from traditional drawing to digital – and we were ok with that until those innovations were mere tools. This is different, though, and dirtier. Would it be right to talk of AI piracy here? Drawing AIs, after all, use databases containing billions of public images to produce their output. The images were included in these databases without any knowledge or consent from the original authors. Why? Because it was more convenient to do so, rather than ask permission or, even less conveniently, build up a system engineered to be supportive and fair towards artists from the getgo. Programmers knew what they wanted and they just took it.

A few months ago, I was listening to a podcast where a dear friend of mine, the professional Italian vocal artist Edoardo Stoppacciaro, was denouncing the rise of algorithms who can imitate a dubber’s voice. It was quite enlightening. Basically, he said that an AI can reproduce the voice of an actor; sometimes the results were great, sometimes terrible, but it couldn’t imitate the subtle expertise or even the tiny mistakes of a human being that make the acting feel real. His point, though, was that voice AIs were risking to put the entire category of voice actors out of business, or at least in serious distress. That really hit the spot for me.

After listening to his interview, I started thinking instead of what the positive sides of such a wondrous technology could be when used properly. It could be an aid to voice actors, I inferred. Let’s say a voice actor needs to do more roles than he has time for, or wants to take a leave; the AI could fill in for him. Meaning, the studio could rent a virtual version of you who can voice act, and pay you for it. Nice. What if the voice actor is ill, or he loses his voice? The algorithm could replace him to finish the last episode of our favorite TV series. Even when he retires, or dies, his voice could be kept on acting in new shows by the AI that sampled his data. Imagine having voices like Ferruccio Amendola or Tonino Accolla back, just to name a few fallen stars among the Italian voice actors pantheon! That would raise some profound ethical and economical matters, wouldn’t it? Would their families and fans even agree with bringing their loved ones back on stage, even if they’re not among us anymore? How should their families be paid, what rights would they have? That does remind me of dead actors brought back to life by CGI in Star Wars movies. I found it super creepy, how real it felt. Is that even right? And think of Bruce Willis, who, apparently, sold his CGI image so that he can be kept on acting after his retirement. So, this actor/computer replacement is already happening and could become common in the near future. Such usages are still only regulated by private agreements. However, more than contracts and private agreements, should there be laws that regulate the immaterial projection of an artist’s creativity, such as his/her own image, voice, or indirect artistic creation?

We’ve had such a law since 1790: it’s called Copyright Law, and we all know how it works. Creator-owned content cannot be reproduced without the author’s consent and an agreed amount of money must be corresponded to the original artist if you are, in any way, making profit out of its reproduction and sale. Problem is, the old concept of copyright is showing all its limits in the digital age, and it needs to be rethought and extended to meet machine learning technology. With digital content now being available worldwide, infinitely reproducible and even usable by AIs as base material to produce new content, who’s to say what belongs to who? This legal gap makes creator content prone to abuse.

We’ve already seen that happen with new technologies. First, there’s the stage where enthusiasm and rejection mix up, then comes the abuse, and finally regulations kick in to achieve some form of balance, but not before a lot of irreparable damage has been done. This AI thing with pictures is music and video streaming happening all over again, but this time the story could turn out much darker for the creatives, unless we’re careful and decide to act.

Remember when digital music first came out? People over thirty should. If you don’t, go watch “The Playlist” on Netflix: it’s the story of how Spotify was founded in 2006. Back then, Napster and The Pirate Bay had put the music industry to the spot. Everyone was downloading MP3s from the Internet for free. It felt like freedom, and in some way there was a form of popular justice to it, but in fact, piracy was harming workers of music companies and musicians. We had a hard time understanding it, back then; digital goods all seemed so… immaterial. The music industry was collapsing after decades of total domination from record labels, and that brought no benefit to the creators who fans professed to love. The music industry fought the phenomenon by trying to repress it, to no avail. People wanted cheap digital music and enjoyed carrying millions of songs in their pocket. They liked to get easy access to what they wanted, now that the technical means to satisfy that need were there. The industry only survived by no longer fighting the change, but adapting to it. Thus, a new distribution system was built that made music streaming more convenient than piracy. By paying a cheap monthly fee, people could listen to all of their favorite music. Why bother downloading MP3s anymore? Why spend time searching for songs, ordering music libraries, uploading them to your device a maybe get a malware or two on the road when you could have so much more, and so much more easily, just by paying a teeny tiny subscription of 10 bits a month? Everyone wins. Ok, maybe the creators weren’t quite so happy in the end because they got pathetic royalties, and that’s a problem. But in theory, the model is valid.

The same thing happened to movies and tv series on a bigger scale. Remember when it took three days to download a pirated movie from Napster or Emule? Well I’m grateful we have streaming services now, because I have all the movies I want at my fingertips, on all my devices, everywhere I go. If I want to watch a movie on my phone during my 20 minutes on the bus, I can do it.

Fatal flaws of the digital illustration sector

So we’ve done this transition with musicians and directors. Why can’t we do the same with drawing artists? Well, there are some fundamental differences this time.

All that I’ve said before comes down to a few simple facts:

1) AI as is currently conceived is NOT a distribution channel for human work, but an autonomous system meant as a replacement for human work that doesn’t take the interests of creators into account, thus causing economical and intellectual damage.

2) No effective AI trading system can be designed unless the rights of an intellectual work are both properly defined and easily and clearly traceable via widespread software standards.

3) Updating copyright laws with of a proper set of rules that regulate the economic and intellectual rights of the artists in the field of machine learning is needed.

4) The existence of strong groups of interest who can exert pressure for the application of such principles is a major factor for success in achieving balance.

Unfortunately, none of the above is easily applicable to drawing artists.

The first difference is, AIs are not meant to distribute creator content, but to acquire, rework and resell creator content via a software platform that actively replaces the role of an artist. That means creators are indirectly damaged by it; those who say that it’s a mere tool are simply failing to grasp the problem. If something, human artists are tools for training AI algorithms here! So, the first order of business to get out of this jungle would be to rethink the laws of intellectual property and give more control to the creators against the unauthorized use of their art by AIs. Creatives should be legally enabled to authorize, or at the very least they should have the ability to deny, the handling of their images by AI drawing softwares (on this matter, I suggest you read Francesca Urbinati’s intelligent proposal for a “AI-yes” enabling tag, which you can find here; although that would be far from enough, because it could easily be ignored or worked around).

In order to do that, we would need a world-adopted standard to authenticate images. We have NFT, a blockchain technology, but that really doesn’t work for many people, I won’t get into the details why. The system should be recognized by state laws. Whenever you, as an artist, release your artwork on the web, you do so under a serious legal protection. It means you can sue the ass out of anyone who abuses your work, AIs included. Craking this on a technical level is already a very difficult challenge, if not impossible.

In theory, though, that would allow both control and tracing of digital art as an intellectual property, but it would still not be enough. You can’t defeat AIs this way because they digest and rework your data to the point they become completely unrecognizable. So, we need laws that establish that AI databases can only contain authorized images, and legally ban the use of AIs who do not abide by this rule. That would lay the groundwork to build a participated system where, when a creator’s images are used by an algorithm and the result is downloaded, the creator who has registered as a supplier earns some royalties. So, a creator that used to do 10 images a month by his handiwork, can instead make 5 and keep his standard fee while the AI sells, like, 500 computer-generated images and pays him a small fee for each one. THAT would make AI interesting and not so damaging for creative artists.

Making that happen would not be easy, though. Here comes in another big difference from the previously mentioned digital media.

While the music and movie industry had very big actors in play when the XXI^st century happened, the illustration industry doesn’t. Well, not really.

There are big agencies for illustrators out there, that’s true, but they may not have the power to exert sufficient political pressure to build such a complex legal structure. Besides, the rights of illustrators have never been a priority in anyone’s political agenda, because there has never been a mass pressure to recognize and enforce them. Paradoxically, it was the presence of major companies who played a key role in governing the market transition of music and video from physical to digital media. The digital market was born wild, but the big companies had enough economical and political power to shape its transformation in a way that was beneficial to them and to their customers – not so much to the creators, but still better than a lawless market headed for collapse. In the end, God forgive me for praising capitalism, having a powerful oligarchy who held the cards turned out to be a good thing for everybody. This time.

This is much less true for the illustration market. There has never been a big player so strong, a presence so authoritative to impose an universal creation and distribution standard in the world. Drawing artists have always remained a more independent and fragmented category, its members often unassociated with one another. In this kind of situation, this is a disadvantage. The wave of AIs is met by no united front to match them in strength. Thus, artists may very well lose this competition, where musicians and videomakers have sort of won the challenge of digital distribution.

How this is gonna go down, and what can we do about it

AI is happening, wether we want it or not. However, I think it’s safe to say that AI will not be allowed to go on without a rule forever. The question is when these rules will come and what form they will take. And, what role we can play in writing them.

As I said earlier, we’re still in a stage where AIs are running rampant. It will get more abusive before it gets any better. I do not think that AIs will wipe out human artists, just like TV and ebooks didn’t kill physical books, or to remain more in the topic, digital drawing didn’t kill traditional drawing. But there will be pain and adjustments to make. I’m sure many artists will see an important reduction in their work as companies and studios start turning to algorithms to produce their content for a far cheaper price. Most likely, some artists will start using AI on their own to speed up their creative process and sell the content to companies themselves, but it will take some time before law is brought to the lawless.

What can we do about it in the meantime?

The first thing is, we should NOT ignore the rise of drawing AIs. Rejection is an understandable reaction, but like the music industry taught us, denying progress does more harm than good and it’s the best way to get caught unprepared by the advancing new. Rather, we should learn how AIs work, know what they can and cannot do, so we can understand what competences we need to acquire in order to reposition ourselves as professionals in a shifting market. It is likely that the illustration market in the future will become even more competitive and demanding in terms of digital and technical skills; artists who ignore this could become marginal. Developing a personal style and brand, one that cannot be easily replicated by algorithms, and forging strong bonds of trust with clients will become even more important than how it was in the past.

Another idea is to make AI work in our favor. We could imagine artists supporting each other by building a community around an AI engine that makes an ethical use of their work. How could this business model work? First, we lay the groundwork for an AI-based social platform to which artists can register. Subscribers contribute to the platform by manually uploading their artwork to the AI’s database or tagging it so that it is registered, thus helping to build a common library made up from the contributions of free and willing participants. The artwork is then used by the users via the AI to produce images that any community member can use for its work, but every time an image is rendered and downloaded, royalties are paid to the authors of the source art as a form of passive income. Goes by itself that uploaded images can be removed at will from the database. The system remains under the control of the artists and is an ethical way of doing art business with AI. Think of how this could speed up and improve your individual work, without harming any unwilling artists.

Keep in mind that people understand the difference between an item, or a piece, that has been handcrafted by a human being for them and one that has been vomited out by a machine. The work of a machine can be beautiful and impressive, even more so than yours, but it has no original concept behind it, and thus very little immaterial value, which is the perceived value attributed by a customer to a product. This is what you see in furniture industry vs artisan and design craft. You do not see a hundred woodcrafters in a town today, but a couple of highly specialized workshops visited by selected customers who desire unique items; if you want cheap and generic stuff, you go to IKEA. Man-made work has, and always will have a special value to it, and many people still appreciate it. So artists are by no means going to disappear. But they will have to accept the challenge that has been cast and find a way to coexist with AIs until balance is found.

That doesn’t mean we should simply suffer and wait. In order to sort this out, it is indispensable to stick together and make pressure on AI creators, publishers, artist agencies and in turn, governments to regulate a matter that is starting to impact not just art but more and more sectors of creativity and entertainment. Users on Artstation and other big social networks are already on a full-scale strike against AI-generated images, and both China and the European Government (to my surprise, I admit it) are working on new regulations towards machine learning technology (although it’s still unclear if or how these regulations will involve text-to-picture AIs). Social networks can effectively replace big players by allowing artists to organize and take action, and some initiatives are already taking off; Karla Ortiz, board member of the Concept Artist Association, has started a crowdfunding campaign on Gofundme to promote a class action against AI creators that has reached over $115.000 in just three days. We still have to see how the great manufacturers of hardware and software for artists (such as Wacom, Adobe, Apple) will react when faced with a potentially severe damage to their business income due to this technology, since they could have all the interest in offering financial support to their protesting users. We need to find ways to make ourselves be heard, to make people understand that unregulated AIs hurt creators who depend on their art to make a living. Today AIs have learned to draw, but tomorrow they will make videos, music and dub media. I, as a writer, cannot sleep tight either because there are AIs who are learning to write books. So, unless we want to silently be replaced by machines, we really should DEMAND proper laws and pursue class actions, instead of repeating the same mistakes over and over! The fate of genuine art in many forms depends on it.