The Future of Artists With the Increasing Accessibility of Text to Image Generating AIs Like Dall·e

ログイン/ 登録

A few days ago, I received access to OpenAI’s text to image generation AI DALL·E. It’s amazing, just like I expected it to be. After playing around with it and having numerous discussions with people about it in the last few days, I have come up with some observations which made me develop a few future visions for the role of AI in art creation process and its relationship with the human artist. My experience with text to image generating AI is limited to DALL·E and other mainstream products, but my concepts can be applied to generative AI as a whole. For reasons of convenience, I will refer to it as AI in this essay.

In most conversations, the first question that popped up was whether or not AI such as DALL·E will make artists obsolete, exposing the existential threat that some people experience when they first encounter something new and unexplored. This essay is an experience based projection with the means of answering this very question.

I will first define the different roles of artists in a sociological context and then make connections to how the economical implications that generative AI bring along could influence the role of the artist, both sociologically and economically. Through this I hope to break down different aspects of the artist role and discuss short-term futures predictions, feasibility, but also long-term consequences of generative AI. I will include notions from to scientific papers and books which you can find the bibliography and reading recommendation section at the end of this essay.

What does an artist do?
The artist is a person who, in the most basic way of explanation, creates art. So in order to properly answer this question, we would have to first answer the question of what art is, which was lastly contested by Marcel Duchamp as a pioneer of the Dada movement. Since this essay is about the relationship of artist and AI, I will reserve this discussion for future articles. Art in this essay is used synonymously with words such as “artwork”, “art object”, “artistic creation” or “artistic product”. Below are some of the key notions of the artist role in our current societal and economic setup.

The artist as the craftsman: At its heart, the artist is a craftsman who pursuits the mastery of the skill to use material and mediums and to transform it into something of higher, for example aesthetic, value. During this process, hours go into the practice and study of original prototypes. Developing this skill, the muscle memory, is what a lot of artists take pride in. Adding to this, the artist explores and plays with the learnings of the studies to develop their own ways of abstraction, combination and collaboration. In this sense, the artist can also be seen as a creator of aesthetics. Artists use the given tools and materials and build something that finds appreciation by fellow humans. The artist develops a knowledge of the different languages of aesthetics by studying and understanding humans’ way of perceiving and appreciating aesthetics. Through the craftsmanship, the artist can also use this ability to provide services, for example commissioned works.
The artist captures universal human emotion: According to the notions of Susanne Lange, the artist also captures universal human emotion. (1) Through abstraction, the artist manifests this experienced universal emotion, through mediums. Contrary to the misunderstood belief that artists draw what they make up in their imagination, their role in the art creation process is to observe human behaviour and thought and capture what is beyond the rational, but what is of aesthetic and emotional nature. This explains why many abstract artworks are so popular with the mainstream masses. Ultimately, the artist can be seen as a medium, translating observations of emotion into a visually recognisable aesthetic. The artist as a storyteller naturally produces “visionary models” or “critical commentary”(2) on cultural, political or aesthetic concerns to inspire and provoke independent thinking in other members of society.

The artist as a designer: The artist as a design can be seen in various fields of specialisations. This particular artist is more concerned with the bigger picture of art creation. In this category are for example the directors, the compositors, the curators, the experience and product designers. In order to fit this category of artist, a combination of the two above mentioned archetype skillsets is required. The designer has a clear vision of the end product, the artwork, and has abilities to select, combine and communicate with artists that produce and iterate the smaller parts of the big picture in the most efficient way. Artists of this caliber create franchises, explore Transmedia worldbuilding and storytelling.
The artist as researcher and educator: Leonardo Da Vinci is the best example for the beautiful connection between art and science. In terms of abstraction, art and science have a similar approach. The artist and scientist share the skill of observation, interpretation and execute upon the learned knowledge. Children learn societal values through picture books or animated series co-created through multiple artists. For the artist in the educational sector, skills like collaboration and communication with interdisciplinary colleagues in the creation of educational content are required. In the polymediated world that we live in, the media that we consume or engage with induces feelings in us or can serve as simulation experiences. In the field of research, an affinity for technology is required for the artist. The process of experience design, in for example new media art, is a part of the complexity theory and takes a new approach to science, replacing positivism, which underlies reductionism, determinism and objective knowledge with the fundamental assumption that all knowledge is subjective (3). The iterative process of experience design that complexity theory underlies can be observed in the approach that new media artists use to explore technological and societal potentials of technologies. In data science researchers find aesthetic ways to convey knowledge. Stories have been told since the emergence of humankind in many different ways to teach the wisdom of our species.

Surely there are more archetypes of artists then the ones that I am mentioning here in this essay. Interestingly these categories can also be observed as ways in which the artist is compensated in our free market. Before I draw out the economical implications generative AI has, I will analyse which of the above mentioned tasks of artists can already be done better by art creation AIs like DALL·E.

Which tasks does the AI already do better?
When discussing the future of artists in the scenario of increasing accessibility of generative AI, the next step is to analyse whether AI can actually do the things that artists to now, and if they can do it better. I will use the above established artist archetypes and use deduction based on my personal experience with generative AIs but also my professional experience from working in leading roles in the creative industry.

The AI as the craftsman: In terms of mastery, DALL·E is already more advanced than many human artists. The amount of input information, the “original prototypes” (4) that generative AI learn and can access at any time, is undoubtly much higher than the average artist can even imagine to study in their life time. Additionally, some artists eventually find their own style that they are comfortable with and that they find aesthetically pleasing, so they will also stop to learning new ways of abstractions and new mediums. When asked to transfer an existing style to an image reference, the AI can achieve this data translation within seconds, while a human has to manually learn, build muscle memory and then finally attempt to do something in the similar style.An Example for style transfer is https://deepart.io. This type of style transfer can also be considered more accurate, since it takes into account of all existing data and patterns that it has learned from, and keeps on learning from, while the artist usually only uses a couple of visual references. The speed of an AI craftsman is completely out of the league for a human competitor. Generative AI is still at the early stage and the craftsmanship quality varies a lot with different AIs, however, it possesses computational brain capacity. It is not perfect, yet.
The AI captures universal human emotion: When debating AI replacing the human artist, the question of creativity as something that is only gifted to humans, comes up. Since creativity is such a broad and inclusive world, I believe that the question can more be understood with, is an AI able to capture what is not of rational, but of emotional and aesthetic nature? I think that the answer to this question lays in the very nature of AI. AI needs to learn and the “original prototypes” that we as humans put in. Any of our existing artworks captures our universal human emotion. The AI will probably even notice ways of abstraction that we haven’t discovered ourselves, but are ever present in artistic work throughout humanity’s existence. The AI definitely can imitate and transfer past universal human emotions through their creations. While playing with the creative mode of the Snowpixel text to image generator, I was personally convinced that the generator was “creative” in its way to visualise concepts and making connections between relevant topics.
The AI as a designer: As of this moment, the AI is not truly conscious and therefore doesn’t have its own purpose or will of creation. It can combine different elements and remix them without any problem, but the context of it has to be input by the user. Here again, the AI can perfectly mimic existing styles of large scale productions, but the initiation still lies by the human. With the age of conscious AI, this aspect would change. But then everything would change I suppose and there will be no use of this essay.
The AI as researcher and educator: One of the defining factors for true artificial intelligence is its own ability to learn and seek knowledge and improvement by itself. At this moment, the knowledge base of AIs is already bigger then anyone of us as humans could comprehend. The processing power and connection building can in theory be constant and infinite. With better improvement in this area, we can expect that AI will be able to teach us, through for example simulation. We could consult AIs for complex scenarios, visualisations and experience simulations, broken down so that humans can understand them. Imagine “Talk to Books” but charming, literate and with consciousness. It has the potential to revolutionise individual education.

After this analysis, I believe that generative AI have already exceeded human artists in many aspects. With this in mind, the threat that it opposes in our identities as artists and creators is understandable. The questions of economic impact with being made redundant has to be answered by leaders and solved in a sustainable manner. To propose an optimistic outlook for this current lack of solution, I would like to explore the feasibility of generative AI and how the artists and AI could possibly have a productive relationship with each other.

The new role of the artist and the feasibility of AI
Until our lives are completely taken over by AI in the future, the artist, as all other human beings, will continue to exist as part of society and the economy. The AIs are growing in this very moment you are reading this essay. We can agree on that it is powerful and has huge potential. As history showed, with the emergence of new humanly controllable mediums, the relationship to the medium itself changed for all of humanity. New norms and classification will be discussed and art will still be a great way to explore the mediums. Additionally, because our relationship with the new medium changes, we as human change as a consequence of it. Theories discuss the idea that we have already evolved from homo sapiens to homo techno. So how could the closer future role of the artist look like? What would the techno-human artist create? The next part is about what I envision in my dreams, hopes and wishes for the future relationship between AI and humans.

I think the first change will be and can already be seen in human artist as the craftsman. But it will not disappear. The decrease can be traced back to the advancements in generative AIs and their financial feasibilities, that I will detail out below. There are already massive markets existing that value the human nature of craftsmanship as its main priority. Etsy or Masterworks are great examples from the art market that has benefited from technological advancements and globalisation, while keeping the value of manual skills. The part that I see being taken over by AIs will mostly be in the commercial industry. Many artists already make use of better tools like algorithms and automation on their artwork creation. With better tools, the muscle work that the artists has to contribute lessens. The advantages that AI present in terms of imitation and recombination is of no comparison to the human abilities in this area.

As of now, the one thing that AI can’t replace is the purpose of a creation and therefore the artist’s vision of what is worth manifesting through a medium. Giving the combination of elements meaning. The best generator that you could imagine would still need at least a click of a button, a will of creation to execute its function. And here is where I see the artist shifting towards in the near future. Creation and production will be much faster, with the artist as the designer archetype. Imagine a world where we have AI, that can, upon the spoken world, manifest whole universes in virtual reality. In this type of world, the artist has the power to create parallel universes in the virtual space, with its own physics, species and any type of rules that they can think of through their imagination. For the near future, I see that the artists skills will need to shift towards the ability to communicate and collaborate with AIs, the ability to curate and combine and worldbuilding skills. There are already examples made by contemporary artists like markinducil who created this little concept for Instagram with the help of AI. In this context, the focus of the artist is not the creation of their own abstraction and challenges the assumption that art needs to be completely original. The skill of combination, curation and collaboration is of higher importance for this type of artist archetype. Many attempts have been made to recognise patterns in existing artworks, trying to categorise and classify art. The documentary “Everything is a Remix” gives some examples from pop culture. Claire Bishop argues that even though many artists use digital technology for their creation, they hold onto their “fascination with analog media” (5) to the extent of mimicking analogue mediums, instead of creating from scratch with the new technologies. I believe that we have yet to discover the possibilities of digital art with the help of technological advancements, like AI, before we can purposefully use them to create art from scratch. At the moment, we are still learning and even developing the very tools for future creations. Similar to classical artists like Picasso, who had first learned how to photo realistically translate what he saw with his eyes into a painting, before breaking free from the previously learned boundaries, producing his most well-known abstract paintings. This ties into the another hypothesis that Bishop raises about the role of the individual participation of collection and archiving as a consequence of digitalisation. She claims that with the widespread adaptation of technology usage, every person can combine mediums like film, audio and text, participating in a new form of interactive spectatorship and communication. I would even go a step further and state that this collective act of creative archiving, as a form of historical and social documentation, is merely a necessity to prepare for what will be the next step of art: parallel universe storytelling.

Okay, let’s come down again. Big visions need realistic check ups. When talking about feasibility of AI in comparison to artists as we know them, there are two other factors that play an important role: accessibility and financial models.

Let’s start with the financial models. As of now I have come across three different types. The first one is the payed model, for example the one that Snowpixel uses. The pricing is simple: $10 for 10 credits (100 generated images). If we allow the quality of the image generator to improve over time (which will be inevitable as the AI continues to learn), while the price stays the same, paying 10 cents for generating an artwork through text prompt seems pretty reasonable, since it would take a human artist potentially hours of work to produce the same art through manual labour and craftsmanship. The rendering with Snowpixel takes up to a couple of hours, but because you can submit multiple prompts simultaneously, the user can win back the time through a functioning resource strategy. While I was using DALL·E to create an artwork for the DAO that I am working for, I came across the situation of completely scraping one artwork, because it only took a small amount of time to produce with the help of AI. The speed and mass of art generating AI is out of this world, and its possibility to create infinite combinations, would make the creation of a suitable artwork more resource efficient. Of course the price still has to adjust in the still very young market, but the already existing examples don’t render the impression that services like text to image generation will be an exclusive luxury good. The second financial model is open source and free for everyone to use. Like other open source software, there are many attempts to make free usable text to image generators, for example through the Disco Diffusion Google Colab. Here also, there is also still room for the AI to learn, but its abilities are already impressive. The third model that I could see is a mix of the paid and free to use possibilities. DALL·E for example is funded by OpenAI, which is an artificial intelligence research laboratory consisting of the for-profit corporation OpenAI LP and its parent company, the non-profit OpenAI Inc. Free to use mini versions of like DALL·E mini by Hugging Face already exist. This mixed model could allow each account a certain amount of free generations per time period. With industry and, I would dare to say, societal changing open-source software like Blender as role models, the attempts to make a powerful AI available free for everyone to use are already in the making. OpenAI explained to the artists that were granted access to DALL·E with the limitation of 50 prompts à 6 generations per 24 hours. In the introduction event they have said that there is no plans to remove this access in the future. Of course a text to image generative AI does not compare to the computing power to create virtual universes, but it definitely can change many artists lives, and help them to be much more efficient in their art creation process.

When analysing the accessibility of AI, the computation of the image generation is device independent and can be used through any screen based device with a browser. This accessibility is partly due to a shift to convergence culture, in which less and less devices are needed to be an existing entity in the creation and consumption of art. It allows new ways of artistic collaboration that we didn’t imagine before. If we are capable of implementing AI and automation into other areas of our life, we are looking forward to a bright future, in which humans can take a step back from manual labour and utilise our brains to design, create and procreate. However, we have to solve the problem of energy creation, since AI require power and the digital world for existence.

The future relationship of artists and AI
I hope with this essay I was able to capture my bright outlook for the relationship between artists and AI. One of the greatest strength of humanity is the ability to collaborate and co-create. Another one of the greatest strength of humanity is adaptability. We have overcome and built and improved ourselves to the point where we have discussions about artificial intelligence, which can be understood as the ultimate augmentation of our brains. Because art, as something so closely related to the essence of human nature, wants to be considered as exclusive for humans to create, we take a sceptic stance on AI artists. I have explained different notions of artist archetypes and compared them with and through my analysis i conclude that generative AI will definitely be implemented much more in our lives, taking on big chunks of our manual work. I think the artists’ angst of being made redundant by AI should rather be seen as a gift, an opportunity to augment our creations and expand our imaginations. The collaboration with something that will be so much wiser and more powerful than we are, could be just the beginning of art. Whatever the future may bring, I believe and I hope there will be a mutual camaraderie between the human artist and AI.

Bibliography and recommended reading
(1) Langer, S. (1948). Philosophy in a New Key: A Study in the Symbolism of Reason, Rite, and Art. THE NEW AMERICAN LIBRARY.

(3) Heylighen, F., Cilliers, P., Gershenson C. (2007). Complexity and Philosophy. In Bogg, J. and R. Geyer (eds.) Complexity, Science and Society. Radcliffe Publishing, Oxford.

(4) Geng, Y., Du, X. X., Zhao, A. (2017). New Media Art as Expressive Language — Artistic Abstraction, Illusion, Emotion. 2017 4th International Conference on Social Science (ICSS 2017). Pages 257–261.

(5) Bishop, C. (2012). Whatever Happened to Digital Art? Artforum, September issue. Pages 434–442.

Benjamin, W. (1969). The Work of Art in the Age of Mechanical Reproduction. In Illuminations, ed. Hannah Arendt, New York: Schochen Books.

カタログ