AI Generated Pictures — Follow-up Essay

Jan 25, 2023 | Journal

Since my last blogpost about AI art, I decided to write a follow up on the matter, because there are many things I see differently now. I also understood why a lot of people are against it. So here I am again, and now I would like to offer a new perspective, meshed from two very different, but somehow connected worlds.

I am sure that many of you have seen, the topic of “AI art” has completely taken over the creative community. It was not by accident and it did not surprise me at all. Even I tried to experiment with AI art before, which experiments can be considered light-hearted or even naïve adventures from my part. AI is here to stay whether we like it or not. For some it is an unfortunate for some it is great help. During the last couple of months, I once again talked a lot about this topic with many people from a wide range of backgrounds – from the ordinary person, through university students, my peers and colleagues, to even programmers and experts on the subject. Since I am affected by this as well, I also want to take my stand regarding the matter.

I am not speaking out against AI as a tool, but I felt that the discourse started to go into a wrong direction. Recently more and more people are upset and try to object to the application of AI, which is understandable. So, my standing is more or less against its current implementation and use. Technology is not evil, humans are. Therefore, my main argument revolves around the moral aspect of this issue.

In the field of translation, a lot of translators and offices use AI based translation machines to make their work faster and more efficient. However, this does not mean that, for example, if you use Google Translate your document will be aptly translated. This just means that there are vast amounts of different texts, which are fixed in their meaning, such as legal documents or manuals. However, when it comes to literature, there is a need for human cognition and creativity. In comparison, literature is very flexible, due to its ambiguity, and therefore requires interpretation, while a fixed meaning document is more or less written in a way that it means the same in every context – of course courts my beg to differ, but there is no doubt that the AI can translate a trade agreement better, where a person agrees to sell a certain amount of crops to a company. Also, when it comes to literature, everyone has a style and the AI is only capable of learning and mimicking a style based on the input it gets. Therefore, it cannot come up with original things – which connects my point to the views of artists. My work, just as theirs is in danger, and beyond us, there are a lot of other trades that AI can already do better, if not then it will in the near future.

Many artists use AIs in their studios for edge smoothing, texture generation and basically, for the same reason as us in translation, to speed up other processes. The customers expect from both of us a level of precision in execution, which the AI is nowhere near capable of executing. Nevertheless, I can count myself lucky in this aspect. The spread of AI at this level hurts freelancers much more than those who specialize themselves to the needs of larger clients and are now bound to work for them. Therefore, the fears of freelancers and aspiring individuals new to these fields are completely justified.

For example, those clients, who have previously paid thirty, fifty, eighty or so dollars for a D&D character portrait, can now have it generated with AI or at least more or less have what they wanted. However, those who know exactly what they want and expect a high degree of quality will continue to turn to artists… Or similarly, with the emergence of AI based translation machines, people often turn to Google Translate and such to get what they want without paying. However, it does not work in many context. I often receive job requests, where I am required to correct AI translated texts. In some occasions, they turn out to be mistranslated literary texts. Therefore, those who want quality, just as in the case of artists, will turn to translators instead of translator machines.

The main problem – which I could identify, based on the many discussions I had in the matter – is that some people pretend to be artists with AI-generated images and pretend to be writers or translators doing the same with AI translations or AI written texts. Furthermore, they accept commissions and take payment while basically doing shallow work or outright stealing from other artists and writers… The market has become diluted with fraudsters and so called ‘techbros’, who often times attack creators for fun. For instance, there are examples where some people or groups created a database of someone else’s works and generated images that were not created by the artist, but created in his style by the AI instead. This practice and attitude hurts every industry and creates a general distrust.

However, there is more at stake here than the livelihood of various freelancers or creators, because AI has revealed a rabbit hole where not only moral but also legal problems have arisen. For us creators copyright is a very serious issue, when we aim to live off our works, be it translations, writings, or art. Probably, most people have already heard about copyright, but many advocates of AI-generated artworks deny its significance. This is the part, which nobody pays attention to when building AI databases and the most terrifying is that they purposefully do not want to.

This problem is more significant and more obvious in the field of art than in writing, as of yet… Databases built using legitimate images are incapable of creating artworks in the same quality as the ones built with stolen works from artists that never consent to this. AI needs quality input in order to learn and improve. At this point, many people argue that AI creates a completely new image. Which is of course true from one perspective and false from another. It differs from the images that are in the database; however, there is nothing unique in the created image, it is a combination of aspects taken from the database. In addition, some also argue that it does not matter whether the images in the database are legal or not, based on the previous argument. However, they immediately start to defend themselves as soon as they are asked to remove the non-legal images from the database.

On the one hand, as it can be read on many forums and in some social media posts, the argument so far has been that this is not theft, but piracy! Okay, but when did piracy become a morally positive term for theft? On the other hand, it is focused on the notion that this is about finally allowing the little people to create good pictures for themselves. To a degree, I can agree with the latter argument. However, the advocates of illegally compiled databases use this as a way to validate their theft. The artists should and must have the right to protect their own intellectual and creative property. If it does not matter so much that their work is part of the database, why do they fight against those, who want not remove these pictures?

In addition to the copyright issue, there are licenses that can be used freely, but with several attachments and conditions. For example, what I gathered is that there are databases and websites that require all images that were used to generate the picture to be named, such as the source or the person from whom the image originates; even if it is/was edited or just makes up a little fragment, or if only a part of it was used. Similarly, when I write an academic paper I have to reference everything that I used, no matter why. This way the original author gets the credit and more or less the artworks are protected. In my opinion, this is a step into the right direction.

Furthermore, if the moral and the legal part were not enough, then there are the aforementioned ‘techbros’. Most of them are disillusioned with NFTs, follow an influencer or a millionaire, or switched to AI, but some of them are currently attacking different artists, animators and writers. According to these people, the creators of the old world have become redundant and are working at an unreasonably expensive price in the first place. However, the reality is that most artists, writers and others have been undervalued by everyone for decades and even since the dawn of time. The current situation has pushed the discourse in an even worse direction. For instance, the pictures of artists are saved, then ‘enhanced’ with AI and sent to the original creators as a kind of mean joke, or worse; these are posted by the thieves under their own name.

As I argued before, the so called AI artists accept commissions and throw an image generated with AI to the customer. This on its own would be fine, if the art was to be generated with a legally built database. It is even worse that the customer in most cases does not even know about this. It only takes days to create a database of anyone’s works. When a new artist enters into the field and given that he or she is not careful enough, there is already a database of his or her works and images are already generated in his or her style… At the moment, no creator is protected from this practice – there are but a few regulations on a few sites and certainly, there is no legal protection in the matter.

And now comes a question that I asked from myself multiple times: if you generate an image with AI, supposedly in someone’s style, then accordingly you have used their work… ultimately, whose work will the finished image be?

Some artists argue that the phrase ‘AI art’ is deceptively bad. This is why tech bros call themselves ‘AI-artists’ and talk about AI as if it were their brush… Here I have to confess that when I naïvely tried to create art with AI, I had the same feeling. I even wrote about my experience here. However, after discussing the matter with others and reading about the matter in more depth, I have to reconsider and ask different questions regarding the moral and legal background of such things. For example, if I ask my pen to write a story, nothing will happen. Similarly, if an artist asks the brush to make a painting based on specified criteria and carefully phrased commands – that is how some AI art generation work – nothing happens. It requires skill to create something unique, and eventually only a human author can give life to it.

In summary, the AI generates the image based on a database and not the user. The relationship between AI and man can be an interesting experience, but it soon becomes soulless and dull. It does not reflect the relationship between tool and artist at all; instead, it resembles a money centred contract between an operator and a client. Human creativity cannot be fully exercised through such means, it requires time and dedication. Moreover, all the credit for the finished picture goes to the programmers who created the AI and not to the one who wrote the commands. Therefore, this problem is very complex. It is important to outline that most creators do not want AI to be destroyed, we only want a regulation that protects our intellectual property. Ultimately, copyright is a central piece in the matter with is legal and moral aspects.