AI and Art: a dream to some, a night­mare to others

Lene Marie Ortega Sæthre

When it comes to AI, it can be difficult to separate fact from fiction. The term Artificial Intelligence carries with it strong conceptual connotations after years of use within the entertainment industry. AI has been portrayed in a myriad of different ways across various genres and platforms, from films like the Matrix, to game like Halo. As seen in the film A.I. Artificial Intelligence from 2001, we often see AI portrayed as a reflection of our humanity. Even when it is not, AI is often shown as autonomous—a thinking entity with its own opinions of the world around it. Reality, however, rarely lives up to fiction, which might be why the topic is so easily disregarded when mentioned within the public sphere as a reason for contemplation or concern; we humans seem to find it hard to set aside the connotations of fiction for the facts of reality. And the fact is: AI is already an integral part of our daily lives. AI are in cars, social media algorithms, surveillance systems, and many other places and systems. And our lack of understanding of this fact may leave us, and our futures, vulnerable in ways we cannot foresee. This is becoming all too clear in the ongoing discourse within the creative industry after the recent advent and release of publicly available image generating AI.

The rapid development of AI is now evident within the creative industry. When studying illustration in 2015, it never occurred to me that AI would, within the near future, be able to create art. The thought of AI being of any real consequence, or threat, to creatives was as alien as its depiction in fiction. It seemed to me that the consensus during my study years was that AI at best could create amalgamations of existing imagery, not create pieces from scratch on its own. Though the first image generating AI was announced a couple of years back my views of the future did not change. Looking back, even if only a couple of months, the attitude that AI is of no threat to creatives seems to me to have been a prevailing one and to have remained so until recently.

Since the various AI image generators were released to the public, this earlier attitude towards their potential impact on the creative industries has changed drastically. With their open-source approach, generators like Stable Diffusion seem to have have rapidly gained a large following, and other AI image generators based on the same source code have begun to crop up at a rapid pace. Only a couple of months have passed since the release of Stable Diffusion, which has since led to a growing malcontent amongst people in the visual creative fields. The concern is not the AI mode of function, nor is it about the potential of what these AI could become and do. No, it is malcontent due to why they came to exist at all.

animert gif AI

Frail egos or something more important?

I do not think artists, myself included, see an inherent problem with the idea of image generating AI. Granted, there are aspects to them that can feel threatening. Especially if you have spent years honing a craft you fear might become redundant. Even so, I admit that they do offer some new and fascinating opportunities for idea generation and iteration.

Anyone can use these generators to ‘create’ visually striking imagery. On their own, however, that is all AI can do, as of now. Yes; you can create a narrative through thoughtfully crafted knowledgeable prompts. Yet, the lack control over the process means that should you want, or need, a clear and concise piece of visual communication then you still need the understanding and the skill to alter the images generated by the AI after the fact. This becomes especially true when taking visual semiotics and its importance in visual communication into account. Ultimately, to get the most out of image generating AI, you would ideally pair them with good artists, which can make the technology slightly less intimidating.

Perhaps it is this understanding of conscious decision-making that lulled artists into a false sense of security in the first place. Why so many creatives were hesitant to be openly critical to this new way of ‘creating’ imagery. As artists, should we not welcome new tools that might expand upon our abilities?

What has changed?

The reality is that nothing has changed, at the same time everything has changed. Personally, my view of image generating AI and their potential is the same now as it was when they were first released. AI can be a seriously fascinating tool for idea exploration to a degree which we have yet to fully comprehend. And there is a valid worry as to jobs being lost to automatization. AI still feels to me just like a tool which can lead to a net gain rather than a loss. If handled with care.

Learning, not programming

There is one thing we need to make clear in order to discuss AI and its capabilities. AI is not directly programmed to do what it does, it is taught how. While more traditional programming relies on input which results in a set response, AI’s behaviour seems more akin to that of a toddler than a computer. Just as a toddler receives information from their surroundings which it then applies through trial and error to, for example, learn how to walk. So does an AI apply information through a process of trial and error to figure out how to perform a given task. We can help it along the way by pointing out desirable results, but overall, it is a process that is in large part autonomous. To kick-start that process an AI initially needs to be fed information to learn from. In the case of image generating AI, the information fed to the AI is selected by actual people. (LAION, 2022) The information in question? Images, and a lot of them.

The images the AIs learned from initially were picked from a variety of online sites by Laion, a non-profit organization aiming to create datasets for machine learning. The fact that they are a non-profit—that they do not download or host images themselves, and that they claim the datasets are created for research—means that they can circumvent the current copyright laws and freely choose which images they want to include in their datasets. As it turns out, one of their datasets, LAION-5B, was used to train the Stable Diffusion AI. After its release, we have become aware that LAION-5B is rife with not only copyrighted works by established and less established artists, but also private photos posted by individuals online (Xiang, 2022). Therefore, many artists have realised their work has been taken by a non-profit organization and used for-profit AI (Vincent, 2022). Because of this, many visual artists are now realising that their signature styles, which many built their careers around, can easily be emulated by AI. For some of the artists concerned this has already passed the point where they find their original works drowned out by the sheer quantity of AI generated imagery made to emulate their style, and often with their names attached to the derivative AI generated works (Heikkilä, 2022). Considering that this is an issue many artists already face, there is really no way to foresee the very real impact this could have on the creative industry in the long run.

When questioned about the use of copyrighted material in for-profit AI, it is common that the companies behind the AI code deflect by pointing to their open-source approach and by citing their research-based neutrality as a justification (Vincent, 2022), despite, for example, Midjourney requiring you to pay to use their services (Midjourney, 2022). At best, there has been some mention of the possibility to opt out of future datasets, which is a solution that might force artists to consistently have to spend time on opting out of datasets, and which forces the artist to constantly police whether others use their content without consent, rather than placing the responsibility on those who feel entitled to other’s styles and even copyrighted works.

Copyright can be a contentious issue for creatives, especially when dealing with breach of it. Even so, it has been a system which most rely upon when pursuing a creative career. To see this system circumvented completely by companies profiting of AI trained on copyrighted material lies at the heart of the problem artists now have; here we find the source of the malcontent which is fuelling the current discourse taking place amongst creatives.

Right now, it is impossible to know where AI will lead. If we look to history for answers, we find that our track record of implementing new technologies in a careful and humane manner to be less than stellar. From factory workers suffering under the Industrial Revolution, to jobs lost to automatization and the subsequent poverty faced by many, the impression is that we too often rely on quick and easy solutions rather than thoughtful and deliberate decision-making. If there is fear now of jobs being lost, then the reason can be understood by looking at history. A discussion around what AI is—what it could be, and how we use it—is called for.

Though I am fascinated with AI generators and their potential, I do see a very real need to have an open conversation on the matter. The creation of Art has been hailed as innately human, and the argument that AI can help anyone to create anything is one that many will find tantalizing. But is the lure of quick and easy images worth it, if it comes at the cost of artists? … at the cost of our own ability to create?
And why do we have a copyright system if it can so easily be disregarded?
What can we do to protect the work of artists from being taken without consent?

Realisation

I had hoped to draw a conclusion about how AI image generation may impact the creative field in the long term. But I find it difficult to draw a conclusion as of yet. Too many questions remain unanswered. I am filled with trepidation when faced with what strikes me as a blatant disregard for the work of living artists and creatives and their work and opinions drowned out at the whim of others. How we think of copyright and its implementation needs rethinking right now. Otherwise, I am afraid we might lose a flourishing art scene.



Refrences:

How Facial Recognition is Changing Society
The Curious Case of Uncanny Valley