A Philosophical and Artistic Perspective on DALL·E2
TLDR: Exploring the Intersection of AI and Art: An intellectual dialogue on integrating DALL·E2 in the creative process
Thanks to Prof. Yaron Senderowicz and Liav Isaac Shopen for their willingness to be interviewed for this blog post.
Intro and Motivation
DALL·E2 has been out for a while, and I thought it could be interesting to write a blog post about it from a philosophical and artistic standpoint. As a data scientist and hobbyist printmaker, I was intrigued to learn how this method is viewed and used for artistic reasons.
While engineers are frequently focused on making things work, it is worth shedding more light on other aspects such as artistic, aesthetic, and social considerations, which often receive less attention. The main motivation behind this exploration was to better understand and explore in what sense DALL·E2 is artistic and possesses human-like qualities. In other words, how should we interpret DALL·E2 output?
Furthermore, there is a growing concern about the impact of DALL·E2 and its consequences for employment. The images it generates are so appealing that it can be difficult to tell if they were generated by a human or an algorithm. For example, we just witnessed an AI-generated photo that won a photography prize, emphasizing the importance of this issue.
In this blog post, I had the honor of interviewing two intriguing people. The first interviewer is Prof. Yaron Senderowicz, a Tel Aviv University philosophy professor recognized for his studies on human consciousness. The second interviewer is Liav Isaac Shopen, a master printmaker, whose artistry and practical experience lend to a unique understanding of DALL·E2 and how to incorporate it into his artistic work.
Interview with Prof. Yaron Senderowicz
Ido: Yaron, would you mind sharing your thoughts on DALL·E2? Do you believe this algorithm can, in certain ways, not only mimic but also display human abilities?
It’s not entirely clear to me that the algorithm authentically replicates human capacities. For instance, humans can automatically create art, but the key question is whether the art has aesthetic value. Criteria can be devised to determine whether or not an artwork has such merit. One criterion, I feel, is how sensitive an artwork is to context.
We exist in a certain culture, period, space, and civilization. We can easily discern contrasts between different cultures. For instance, if a Mozart piece was presented in the 12th century or the 21st, its aesthetic value could vary significantly because it is innately tied to the context in which it was produced since it is inherently linked to the context in which it was created.
Creativity, Meaning and Context
Ido: Creativity is about meaning and how humans perceive it. So how should an artist’s creation be interpreted and assessed?
Yaron: Technology can replicate Van Gogh’s style, but we should question ourselves whether the art we create using DALL·E2 offers something novel or if it’s merely a derivative of existing work. If you have the technological resources to flawlessly replicate Van Gogh’s artistic style, it doesn’t inherently imply that the created piece is art. It merely exhibits impressive imitation capabilities.
One possible approach to understand DALL·E2 is using the concept of ‘Literal Meaning’. Humans have the ability to discern whether a language use is correct, even if it hasn’t been explicitly defined or its conditions of truth explicitly outlined. John Searle, an American philosopher who is widely known for his contributions to the philosophy of language, contests the notion of ‘Literal Meaning’ (which is basically a straightforward and direct interpretation of a sentence, without any figurative or metaphorical elements).
The concept of meaning is intrinsically linked to..well, context. The issue at hand is whether it’s possible to provide a description including the necessary and sufficient conditions for a certain term to apply to a worldly object. Furthermore, whether our comprehension of the definition (or a computer’s comprehension for that matter) will be precise enough to correctly apply or use the term in a conversational context we’ve neither contemplated nor encountered before.
We understand a sentence despite our vocabulary and the extent to which we’ve used the term being finite. The instances of using the term based on memory are definitive and limited. Now, an example that illustrates this involves the definition of the concept ‘on’ in the article titled ‘Literal Meaning.’ Let’s consider a phrase we’d like to understand — “the cat is on the carpet,” or “the cat is on the floor.”
Is there a way to accurately deduce the semantically necessary and sufficient conditions for appropriately associating these terms with a cat and a rug in order to describe such a situation? To affirm that a cat is on a carpet, what fraction of it must be on the carpet? How are we defining that portion? One leg? Three legs? And what if it’s two legs on and two off the carpet?
Can we precisely determine when it would be correct to say the cat is on the carpet? Let’s illustrate further. Imagine flying in space where you see a cat and rug positioned upside down. The cat appears inverted, but who’s on what? Realize how we’ve used the term ‘on’ with the presumption of a background — Earth’s gravity. This unconscious usage merely exemplifies how much context underlies our statements and our comprehension of statements and terms. Our declarations always presuppose a background that is never explicitly stated or fully fleshed out.
To be more precise about the background, consider another scenario where you travel to the United States. You are hungry in the desert so you stop at a diner and order a coke and a hamburger. Suddenly, you’re given a 5-meter hamburger and a 2-meter glass of coke. Is this what you ordered? Probably no, but how did you specify this in your order? Anyone who understands the context — “what it means to buy a hamburger” — knows that this is not what you ordered. And this context obviously varies across cultures.
You might claim that it’s not what you ordered, but countless possible situations cannot be adequately described. We always understand within context.
Human vs algorithm creativity
Ido: What is your view on Human vs algorithm creativity
Yaron: Connecting this discussion to creativity — human creativity manifests itself in the ability to comprehend a creative use of a term that lacks a specified particular definition. Yet, anyone with proficiency in that language will either agree that the new application is correct or they will dismiss the application. Still, there will be a few who dissent regardless. Nothing here is precisely defined and such precision would require significantly richer logical systems to make it clear.
My primary assertion is that we should always probe individuals who believe that they’ve discovered an algorithm, especially one related to art: Did they merely produce an algorithm that successfully imitates a style or work developed long before the algorithm was conceived? We can only answer this question if we showcase a computer that can produce a novel movement in art, or write a new book.
Picasso, an exceptional figurative artist, commonly taught this style before exploring others. Art is layered historically, demanding an understanding of its roots and foundations. When a computer succeeds in initiating a new movement in art — that is, when critics or experts exclaim, “Wow! We have never witnessed anything like this before” — and if this movement genuinely hails from understanding the historical state of art, integrates aptly, and initiates a change, then that is significant. When that happens, and it successfully crafts poetry that redefines how poetry itself is understood, as Borges stated — after Kafka’s works, people suddenly discovered numerous pre-existing Kafkas.
Imitator vs innovator
Ido: What about addressing imitator vs innovator
Yaron: There is undeniably a contrast between being an imitator and developing creatives. One can appreciate the creativity and originality in an artwork
Many developers of these algorithms may argue that there is no such thing as universally accepted innovation. They might think it’s too nebulous or indefinable. True, art is continually being debated. In today’s world, however, academics referencing a modern text can distinguish it from a non-modern text. Hermeneutics (the theory of interpretation) is concerned with this. This is a key concern for anyone attempting to parse meanings between texts. They specifically deal with this type of issue, struggling with the historical character that stems from the cultural context of things
To actually become a part of our lived experience and contribute significantly to it, a computer must evolve from a mere imitator to an innovator and maker. It should aim to produce something new rather than simply replicate what someone else has already invented. I am not claiming that such an algorithm does not exist, but we should be cautious about asserting its presence prematurely. I want to say that the test is not the ability to reproduce or fake someone else’s painting style but rather the ability to make a new cultural shift.
A truly creative computer should generate a new genre. The advent of Cubist painting was initially seen as a derogatory deviation from conventional artwork. Only after our aesthetic adjustments to the works of the Impressionists like Monet, Manet, and Renoir, did we retrospectively appreciate the traces of Impressionism that persisted into the 18th century. Suddenly, it pervaded the public domain. The evolution of an innovation can be likened to a metaphor; it starts as something refreshing and unexpected, and over time, acquires a literal, more defined meaning.
This evolution always bears a relation to a certain temporal, political, social, and cultural context and the way we interpret the past. It’s an intricate process and can’t be depicted by mere software that extracts the rules that Van Gogh used to paint. The act of painting is much more than that. Van Gogh’s works were concrete acts made by a concrete individual who innovated beyond existing norms.
A distinction must be established between technological achievement and creativity. We need to investigate how creativity appears in human life as we know it, and then assess if the algorithmic output matches that.
Interview 2 — Liav Isach Shopen, a Master Printmaker
Ido: Liav, what are your thoughts on DALL·E2 and how we should adopt this type of technology?
Liav: I believe that DALL·E2 should be viewed as a tool rather than a replacement for the artistic process. Begin with an initial idea, formulate a prompt, and see whether the outcome inspires you. DALL·E2 appears to me to be a filter for the concept you have in mind. You can enrich your art by using the product of your prompt as a visual reference point. The idea is to inject your unique artistic style into it while experimenting with the offered image. I manipulate it, continuously seeking variations on the original image I created. The art becomes more accurate as this process is repeated and fresh images aligned with my style are generated.
Instead of sifting through Pinterest or other platforms, DALL·E2 allows you to tailor results to your artistic preferences.. In this regard, I believe it could supplant traditional art platforms by providing artists with personalized references.
Ido: Some may argue that the outcomes of this technology are so beautiful that they could only have been made by humans. What are your thoughts on this?
Liav: Some results are indeed exceptional, yet, they still feel somewhat cold and detached. In some respects, this technology gives the artist back control of their creative because they must extract the reference they want to create, which they may not always be able to do. It is important to note, however, that this reference is not the final product, but rather an intermediary stage in the artistic process.
Ido: How have you used DALL·E2 for your own purposes?
Based on an image I generated, I created a mezzotint print. I prepared a precise word description of what I was seeking for, and this motif assisted me in conceiving an image that was different from those I usually work with. Ultimately, every artist has a set of motifs and this tool can help extend that range. I’m not sure the final product would have been my first pick without DALL·E2, but it did make the development process faster and more diverse..
Ido: What motivated you to use DALL·E2 from the beginning?
Liav: It was partly due to the buzz surrounding the tool, as well as my curiosity about its potential contribution to my creative process. It was less about being impressed by its powers for me and more about figuring out how to use it and incorporate it into my work. DALL·E2 makes art more accessible by incorporating aspects such as finance, time, and ease of effort, which aligns with the current wave of DIY movements..
Using it as it is does not necessarily make you an artist. I think that just generating an image doesn’t warrant an artistic label, an additional and individual layer or stage is required.
Ido: This conversation leads back to my interview with Yaron — does this represent creativity? Does it mimic some aspect of being human?
Liav: I believe that artwork necessitates originality, to create in relation to past creations and within a specific context and language. At the moment, we’re using a language someone else created and making variations of it. Our role as artists is to contribute our perspectives, thereby returning creativity to the artists and generating novelty. In terms of aesthetic attractiveness, I believe it comes down to personal preference. We must eventually question ourselves whether we feel attached to the final product. While I believe that creativity is inherently human, how we measure creativity can’t be applied to computers. Humans frequently project human characteristics onto non-human things, but these should not be mistaken for reality.
Ido: So you’re saying that thinking of DALL·E2 as a human-like ability derives from our worldview?
Absolutely. It’s a common occurrence. Consider abstract painting as an example, where every viewer sees something slightly different. We project what we’re familiar with, what we desire. It’s part of being human. To effectively use DALL·E2, you must invest in the creation process and guide the algorithm, especially when trying to capture something more abstract like a feeling or atmosphere.
Conclusion and Main Takeaways
Art, in essence, addresses the concepts of originality, invention, curiosity, and cultural context. The fundamental question that we should ask ourselves lies in our relationship with the AI tools we use. Do we merely employ them to create an output, or do we infuse our unique and individualistic touches to express ourselves genuinely?
As I see it, the main takeaways are:
(1) While algorithms like DALL·E2 have the remarkable ability to imitate human artistic styles, it is crucial to question whether they truly exhibit the novel and original qualities of human creativity.
(2) True innovation in art requires the ability to generate new genres and initiate cultural shifts, rather than mere replication of existing styles.
(3) DALL·E2 should be seen as a tool to enhance the artistic process, rather than replacing it entirely. It can be used as a visual reference point and filter for ideas, allowing artists to inject their unique style into AI-generated images.
(4) The human tendency to project human characteristics onto non-human things, like DALL·E2, should not be mistaken for reality.
References
Searle, J. R. (1978). Literal meaning. Erkenntnis, 207–224.
