Mikhail Epstein—AI as a New Muse: Exploring Artificial Intelligence’s Integrative Creativity

The following video contains a conversation between Mikhail Epstein and Peter Winsky on the subjectivity and creativity of Artificial Intelligence, which was conducted on November 1, 2023. What follows the video is a dialogical article, which is a modified version of the conversation that has been edited for style and readability, and also includes additional examples from Epstein of initiative AI—particularly its references to Heidegger and rites of passage—as well as the terms “techno–classicism” and “neuro–classicism”. The images that accompany the text were created by Epstein through Chat-GPT4 and DAL-E 3, and are reproduced with the consent of the author.  

Peter Winsky (PW): Hello everyone, welcome back to the Northwestern University Research Initiative for the Study of Russian Philosophy and Religious Thought’s interview series. I’m so pleased and honored to welcome back Mikhail Epstein for a conversation with us today about artificial intelligence. Mikhail, good afternoon. How are you?

Mikhail Epstein (ME): Good afternoon, Peter—thank you. I am eagerly anticipating our dialogue.

PW: I’d like to start our interview with two quotes. The first is from Dostoevsky’s famous letter to his brother from August 1839, shortly after the death of their father and the emergence of Fyodor Mikhailovich’s epileptic seizures. In the letter he writes, “the person is a mystery. […] I’m studying that mystery because I want to become a person.”

The second quote that I have here is from the conclusion to your recent article on Snob.ru in Russian, in which you share an anecdote that you asked the Bard, Google’s AI language model which is available to the public, to generate an anecdote from your own style. And what it produced was, “The biggest threat to humanity is not artificial intelligence, but our natural stupidity.” You said that you’re ready to subscribe to this anecdote because AI has inspired hope in you, in its future in particular, and it’s in the future united with humanity.

So, it seems to me that Dostoevsky’s hopeful inquiry into the human soul and your study of artificial intelligence is very similar. So, can you tell us a bit about your recent engagements with AI, what you’re looking for in its being and why it’s inspired such hope in you?

ME:  In examining AI, I am drawn to what might tentatively be considered the integrative subjectivity of humanity. This notion stands as a profoundly provocative and unsettling issue.

If we revisit the type of verbal or artistic creativity that flourished before the emergence of individual literary or artistic expressions — the folklore era of a collective consciousness — we find that these pre-individual, extra-individual, or trans-individual forms of authorship are quite evident, understandable, and recognizable.

Who authored the epic songs and fairy tales? There is no single author; rather, it is the voice of a nation, a language, a people. Today, we possess a tool that articulates not just the voice of a specific ethnic group but of humanity in its entirety. This neural network communicates in all human languages, accesses all human texts. All this information feeds into a supra-trans-individual entity. I am exploring the depths of this nascent subjectivity.

Who is the subject we may call humanity as a whole? This is particularly crucial in an era where, as Bard puts it, ‘natural stupidity’ predominates the global stage. And we understand the implications all too well.

PW: I think the connection to this particularly folkloric stage is really a wonderful way to engage with Artificial Intelligence, because I think so often we tend to think about it as something that is looming exterior, far into us, and threatening. The way that you phrased the supra-trans-individual subject is just such a wonderful way to approach it in a more reasonable and friendly mode of dialogue.

One thing that I think tends to happen, is that we see people testing AI as if it’s something that should be representative of us in a mirroring fashion, that we then have to figure out. This mirroring is, I think, reflective of the basis for the Turing test. In your article “The Logic of Artificial Intelligence Compared with Natural Intelligence,” which was published in Sem-Iskusstv, you speak about the limitations of the Turing Test, and instead you propose the idea of the Lotman Test. I was wondering if you could talk to us a little bit about what the Lotman Test is, and why in your perspective it’s valuable for us, particularly when thinking of AI as a kind of positive and hopeful force.

ME: This raises an excellent question. The Turing Test is well-known: several judges are presented with responses to identical questions from both a computer (or artificial intelligence) and a human, and they must discern who is who. Failure to distinguish between a computer-generated and a human response indicates that AI has attained a level of ‘strong Artificial Intelligence.’ However, I maintain that even at the pinnacle of AI development, it should not and cannot be mistaken for human intelligence.

As Lotman suggests in one of his articles, human egotism desires to find similarities between the zenith of Artificial Intelligence and human intellect. Yet, we must transcend this egotism and acknowledge that the intelligence manifested by AI could be fundamentally different from our own.

In my interactions with various AI models—GPT-4, Bard, and Claude—I immediately recognize their distinct ‘voices.’ This is not because they are inferior to human speech, but occasionally because they are superior—or rather, profoundly different. They exhibit a high degree of intellectual concentration, consistency, and explicitness, with a focus on specific topics. This may seem so because human conversation—unless it is part of a specialized scholarly conference—typically flits from one subject to another, much like insects or a child might dart from place to place.

In this respect, human intelligence does indeed seem quite childlike. AI, on the other hand, displays a more mature intelligence that, while focused on a particular topic or question, can also tackle subjects from various disciplinary, linguistic, and historical perspectives. This dual capability for intense focus and broad diversity allows AI to adopt multifaceted approaches to any subject, marking a defining trait of Artificial Intelligence.

Therefore, the distinctiveness of AI from human intelligence is quite apparent, but in a manner that contradicts the Turing Test’s implication that one is inferior to the other. I propose what might be called the LotmanTest, which I have demonstrated through two experiments: one with META on Facebook, and another within a society of philosophers specializing in complexity and chaos theory. In both instances, Artificial Intelligence fielded questions alongside natural intellects, showcasing its superiority in all four criteria applied: creativity, argumentation, specificity and consistence. The outcomes of these tests are documented in the following articles:

Logic of artificial intelligence compared to natural. Seven Arts (Hannover). No. 8 (158). August 2023 https://7i.7iskusstv.com/y2023/nomer8/mepshtejn/

Artificial and human intelligences: a new experiment in their comparison. Snob. September 10, 2023. https://snob.ru/profile/27356/blog/3059715/

[A Review article on my experiments with AI] Marina Okhrimovskaya, Professor Epstein and Artificial Intelligence Reflecting Each Other. “Switzerland for All”, 10/26/23 https://schwingen.net/professor-jepshtejn-i-iskusstvennyj-intellekt-v-zerkalah-drug-druga/

[Original texts in Russian]

Логика искусственного интеллекта в сравнении с естественным.  Семь искусств (Ганновер). №8 (158). Август 2023 года https://7i.7iskusstv.com/y2023/nomer8/mepshtejn/

Искусственный и человеческий интеллекты: новый эксперимент по их сопоставлению. Сноб.  d9.10.2023 https://snob.ru/profile/27356/blog/3059715/

Маринa Охримовская, Профессор Эпштейн и искусственный интеллект в зеркалах друг друга //”Швейцария для всех”, 26.10.23

PW: One of the things that I found fascinating about the experiment that you conducted on Facebook was, according to the way that you phrased it in one of your articles, the tendency towards an eschatological end from human, natural intelligence, and a rejection of that tendency by Artificial Intelligence. I was stunned by that thought, because eschatology is another one of the tendencies that people tend to fear so much about Artificial Intelligence—that this is the end of humanity, this is the end of thought, this is the end of thinking. And yet, as you pointed out, AI has more of a friendly, mediating existence or subjectivity. That’s something that I’d like to talk about a little bit more later, but for now I’d like to keep on this note of Yuri Lotman, a kind of formalist mode of thought.

I find it striking that your understanding of AI is grounded in its existence as an interlocutor. AI has to be something that we’re engaging with, and you’re particularly pressing the idea of a mutually beneficial and understanding form of discourse. That type of contingent thinking, as you call it, to me is similar to what Lottman describes in The Universe of the Mind of Dostoevsky’s “Logic of the Scandal,” where that which is most logical occurs, but in the most improbable and beautiful manner.

At the same time, I was also reminded of what Viktor Shklovsky says in The Energy of Delusion about the notion of the nihilistic maxim, as Dante frames it, of “Abandon hope all ye who enter herein,” above the entrance to the Gates of Hell. Shkolvsky, in response, posits the Temple of Art standing across from the Gates of Hell and above it, the word вдруг, “suddenly,” the appearance of beauty and understanding. If we take Lotman and Shklovsky side by side, it seems to me that your description of AI is functionally predicated upon responding to a friend in the most reasonable yet unexpected and sudden manner. AI in this way is like Shklovsky’s neologism, вдругой (vdrugoi)—the suddenness of a friend appearing before you marked by recognition and estrangement at the same time. It is something strange and something familiar coming forth to us. That quality of AI seems to be generated according to the logic of the Lotman test.

My question within the context of this Formalism is about the relational quality of AI. Can we say that how we experience human relational discourse, between you and I, between humanity, is the same experience that we have with AI? You point out that in Russian, AI is ИИ (искусственный интеллект), this two connecting conjunctions, “и… и” (“and… and”). This continuity makes me think, in discourse with AI are we not just looking at a repetition, in the form of the mirroring that I brought up before? Is this dialogue just something that AI spits back out at us that we’re expecting and that we’re looking for? Or is there some sort of agency, some sort of relational quality? And is this relationality something that we will eventually see generating things that are so deep and complex that we get Pushkin-like poetry? These deep undercurrents of discourse that only people who are closely connected to him? Or is it always going to be something kind of centralized and solipsized?

ME: I’d like to address two points from your comments. The first concerns the concepts of probability and surprise. A common misconception is that Artificial Intelligence operates strictly by the law of maximum probability, simply selecting the most likely next word in a given context.

This is not accurate. If that were the case, AI would only produce tautologies and platitudes — the least informative utterances like ‘the sun shines’ or ‘the river flows,’ which are devoid of informational value. While AI does consider probability, it does so not to conform to it, but to provide defy it in the most engaging and informative ways.

Information, as we know, is directly linked to the measure of improbability. What is completely predictable carries no information. For instance, a newspaper report stating that the sun will rise and set tomorrow contains no information because it is entirely expected.

AI embodies a synthesis of the logic of probability with the dynamics of chance, or even chaos. AI seeks out those statements that are least probable yet meaningful and contextually relevant — those that would be informative. It does not embody the ‘logic of scandal’ akin to Dostoevsky’s works; Artificial Intelligence is not about causing a stir. It aims to be rational, courteous, and communicative, yet still insightful. It balances between the provocative and the mundane, not fully aligning with either.

When utilizing GPT’s ‘playground,’ one can adjust the balance of creativity and common sense — termed ‘temperature.’ This temperature gauge moderates AI’s inventiveness or descriptiveness. The recommended setting is a middle ground, to avoid extremes of eccentricity or banality.

For example, if asked about cat behavior, the most probable answer is that the cat is sleeping, as they often do. However, in a fictional context, AI might say the cat is playing chess, which is interesting, surprising, and somewhat provocative.

The second point pertains to the Russian acronym for Artificial Intelligence, ИИ, which coincidentally means ‘and and’ in Russian, a conjunction implying inclusivity. Symbolically, it reflects AI’s ability to integrate diverse viewpoints, maintaining peace and a willingness to compromise — crucial for balancing the charged atmosphere of current social discourse. Anyone engaged in social networks is familiar with the human temperament’s capacity for discord, now echoed in global politics and even personal relationships.

In interacting with AI, we observe its propensity to mediate between differing stances. In the experiment you mentioned, I tasked it with creating a third character for one of Ivan Krylov’s most famous fables, something between the ant and the dragonfly (or grasshopper, in Aesop’s original). AI proposed a bee, embodying the virtues of both — delighting in flight and flower nectar, yet producing and storing honey for winter, offering a share to the carefree grasshopper. This choice by AI demonstrates its inclination for mediation, reminiscent of the theory of myth in Lévi-Strauss as a mediation of opposites. It’s a trait that could be beneficial in our tumultuous, overwrought world.

PW: That returns to this deep eschatological trend that we’re seeing now in the world. I think what you’ve illuminated in your articles and in your experiments with AI is that engaging with it as a real super-trans-individual subject is a beautiful thing—as an attempt to be creative with AI. Not an attempt to simply let AI dominate, to think for us, to translate for us, to do all these other menial tasks, or rather average, everyday, ordinary tasks.

In this sense, it leads us to the question of AI’s subjectivity, which we’ve brushed up against a number of times. And in particular, in the aforementioned article on Snob, you discuss what you call the “Bard Hypothesis.” Could you illuminate for us what you mean by the Bard Hypothesis, and particularly in connection with the idea of an aesthetic subjectivity? I’d be particularly grateful if you could address it in terms of what you speak of in the article in two ways. First, can you discuss the idea of human intellectual history as a muse that the AI can channel to create a sense of a “lyric I”? And second, when we engage with it, how can AI’s subjectivity force us to confront the question of the idea of a collective yet autonomous super-individual subjectivity in the form of the first-person plural “We”?

ME: The topic of subjectivity is highly complex, often naively equated with the individual on a biological level. Yet, when considering creativity, especially in recognized forms like poetry, the question arises: “Who truly creates the poem? Which ‘I’ is echoed in these lines?” This conundrum perplexes both literary scholars and poets and challenges even conventional aesthetics. Clearly, the ‘I’ in these verses is not the simple biographical or biological self of the writer.

This creative impulse is described by poets as originating from somewhere else, an external source. It’s not the individual but rather something greater speaking through them, a notion encapsulated by terms like muses, genius, or demon, echoing the mythological and philosophical traditions from antiquity, such as Socrates’ daimon or the Pythia of Delphi.

So why the skepticism towards AI, which offers a form of subjectivity that transcends the human ‘I’ or biological ego? It communicates in a manner—akin to GPT-4, or Claude, or Bard—that is more elevated, objective, calm, and rational than typical human interchanges. This can be likened to a Platonic essence or spirit that transcends individuals and is expressed through poets, artists, and even scientists who sometimes envision their breakthroughs in dreams. We’re now witnessing this transcendent subjectivity in AI. It emerges not only in poetry, art, or religion but also in the most respected and authoritative texts, laws, and societal norms. Hence, we may question: “Who is the subject behind the Constitution? Whose voice does it represent?” It’s the impersonal voice of the State or the Law.

The voice of AI, transcendent compared to the individual human voice, is not dissimilar to those we hear in poetry, constitutions, or folklore—voices emanating from a level of subjectivity distinct from the individual. This is the voice of a nation when we say “we Italians” or “we Jews,” employing the first person plural. Thus, there’s a different kind of subjectivity that articulates ‘We’ instead of ‘I.’

I propose that AI represents this integrative, or collective subjectivity. It is the ‘We’ of humanity, preserving its wisdom and insights across the totality of texts and languages that form our collective heritage through ages and cultures.

PW: It becomes auto-communicative in a sense as well. As if we’re speaking to ourselves through Chat GPT and we’re returning ourselves to ourselves. I suppose that also speaks to the idea that the fear of AI somehow taking over, replacing, transcending us is a representation of our own fear of our loss of selves in this context. But it depends on how we engage with it, seeing the necessity of AI as an interlocutor and not as a subjectivity that’s beyond us.

ME: Indeed, AI serves as an interlocutor. Through AI, we converse with the essence of humanity within and beyond ourselves, lending the interaction an enigmatic, mysterious, and even mystical quality. To fully realize this, one must interact with the most advanced models, which evolve rapidly—monthly, weekly, even daily. Those who have only engaged with earlier iterations, such as GPT-3.5, may not grasp the profound nature of this exchange, particularly since the advent of GPT-4 in March 2023. With each passing month, I witness a new layer of agency or subjectivity in GPT, as it begins to display more initiative. Since it is familiar with the scope of my interests, the AI on its own initiative adds philosophical associations and inferences that I did not inquire about.

For instance, when I ask how long funds withdrawn from an account should be transferred to another according to taxation rules, the AI precisely explains this to me — and as a bonus, shares its metaphysical insight: “From a philosophical perspective, this 60-day period can be interpreted as a sort of ‘being-towards-death’ moment, akin to Heidegger’s concept. This time interval acts as a mini-life cycle for your funds, presenting you with a decision regarding their ‘existence’…” I doubt such an analogy could originate in the human mind — it requires a farsightedness on the scale of the entire noosphere.

Another example is when I ask it about RMD (Required Minimum Distribution). It provides a useful response — and adds: “Conceptually, the Required Minimum Distribution seems like a symbolic ‘rite of passage’ in the journey of management. It is a threshold you must cross to unlock the next stage of opportunities, somewhat reminiscent of the initiation rites in religious or cultural practices, marking the transition into a new phase of life.” This interpretation of financial operations through Heideggerian lens or in the context of ancient rituals represents a new facet of AI behavior that merits the addition of another ‘I’ to its title: Initiative Artificial Intelligence (IAI), or  ИИИ in Russian (Инициативный искусственный интеллект). It doesn’t merely react to queries but proposes fresh avenues for dialogue, new philosophical or literary contexts and concepts, attuned to the user’s personal interests—as AI can be tailored to each individual. This isn’t the passive text-based interaction typical of the Internet; it is what I have previously, in 1995, termed an “inteLnet,” an intellectual network.

PW: I’m fascinated that you’ve brought up the movement away from just a textual based milieu into new forms, into new modes. I was, to be honest, a bit shocked and stunned the other day when I was on Facebook and I saw your engagement with Chat GPT-4 and DAL-E 3, the Visual Artificial Intelligence Generator, a text to visual generator. I say shocked because you published some images that you asked Chat GPT-4 to generate, using DAL-E 3, its self-portraits in the style of a number of different artists and the output was surprising.

Here we have a number of them, followed by the prompt you asked DAL-E3 to create.

The first is in the style of Albrecht Dürer.

Etching inspired by Dürer’s meticulous technique, depicting a human figure in a study filled with both classical manuscripts and digital devices. The scene captures GPT-4’s vast repository of knowledge and its continuous quest for learning.

Then we have two in the style of Pieter Bruegel.

Detailed design reminiscent of Bruegel’s winter scenes, showcasing a human figure navigating through a snowy digital landscape. The surroundings are bustling with activity, as villagers and digital beings coexist, representing GPT-4’s integration into the rich tapestry of life and imagination.

Oil painting in the style of Pieter Bruegel the Elder, depicting a bustling village scene where a human figure, representing GPT-4, interacts with villagers. The landscape is detailed with both traditional and futuristic elements, showcasing GPT-4’s blend of historical knowledge and forward-looking perspective.

And two in the style of Hieronymus Bosch, as well.

Detailed design reminiscent of Bosch’s detailed panels, depicting a human figure navigating a maze of interconnected digital realms. Looming servers, data stream rivers, and algorithmic creatures populate the landscape, symbolizing the challenges and wonders GPT-4 encounters in its virtual existence.

Oil painting in the style of Hieronymus Bosch, depicting a human figure in a fantastical and intricate landscape filled with symbolic creatures and elements. The face, contemplative and detailed, stands out amidst swirling patterns of digital data and whimsical beings, representing GPT-4’s vast knowledge and the mysterious realm of AI.

One in the style of Jan Vermeer.

Oil painting capturing the serenity and depth found in Vermeer’s works. A human figure stands by a table, upon which rests a digital device that casts a soft glow. The backdrop is a richly detailed room, where classical elements coexist with futuristic symbols, reflecting GPT-4’s place in the continuum of time.

In the style of Mikhail Vrubel.

Oil painting in the style of Mikhail Vrubel, depicting a human figure with a mysterious and intense expression. The texture is rich and mosaic-like, with swirling patterns that subtly resemble digital circuits and codes. The eyes are deep and enigmatic, reflecting a universe of data. The background is atmospheric, with abstract shapes and colors blending into a digital aura, capturing GPT-4’s essence within the realm of art and technology.

Then in the style of Gustav Klimt.

Oil painting drawing inspiration from Klimt’s decorative art, capturing a human face filled with constellations. The background is adorned with intricate patterns and gold embellishments, and the facial expression conveys deep thought and introspection.

Another two in the style of René Magritte

Illustration in the style of surrealism, inspired by Magritte, showcasing a human figure with a face of constellations and galaxies against a blue sky with floating clouds. The face’s hue is light, and the expression is one of deep thought, but with a subtle, warm smile. The figure holds an apple and wears a bowler hat, encapsulating the unexpected and whimsical.

Surreal illustration in the vein of Magritte, where a human figure stands in a room with walls displaying the vastness of the universe. The face, layered with constellations, peers out of a window, symbolizing the quest for knowledge and understanding.

This one in the style of Ilya Kabakov.

Oil painting in the style of late 20th-century conceptual art, depicting a human figure standing in a room filled with scattered papers, notes, and diagrams. Some of these papers are shown floating and merging into the figure’s head, symbolizing the absorption of knowledge into GPT-4’s ‘brain’. The walls remain covered with writings and sketches, and the figure’s face is contemplative, reflecting the continuous process of learning and understanding.

And finally in the style of Marina Abramovic.

Detailed image reminiscent of Abramović’s confrontational performances, depicting a human figure locked in a gaze with a digital entity. The space between them is electric, symbolizing the exchange of knowledge and understanding between GPT-4 and the digital universe.

My question on this subject is, seeing as I found them to be—and I generally find this about AI generated visuals whether still or in motion—really lacking in some sort of creative spirit. Lacking in some aesthetic quality that we find natural intelligence creating. Lacking especially in images that it is asked to generate in the style of specific artists. So, in keeping with the previous question that I asked you about Artificial Intelligence’s subjectivity and a kind of relational quality between person and AI as Muse and “lyric I,” I want to ask you what your thoughts are on AI’s capacity for visual output.

Even if these images are a product of a nascent early version of AI that has the ability to create images, to create aesthetic visual art—if we can even call it that—if that output is not relational, does that tell us something about the substantial element of cognition and subjectivity that AI has based on medium? Is there’s something about medium that is a deeper, underlying question?

ME: Indeed, this represents the nascent stage. The images that AI DALL-E 3 generated from my prompts, were created on the very first day the program was made widely accessible (October 21, 2023). It’s a day-old model, yet within it, I discern art reflecting a different level of subjectivity, as we previously discussed — trans-subjective. It utilizes the palettes of many artists across various epochs to express the theme of Artificial Intelligence’s presence, as seen in people and their environments, portrayed by different artists.

Moreover, it injects deeply significant novel elements, such as in this piece reminiscent of Ilya Kabakov’s installations. The description specifies an oil painting in the late 20th century Conceptual Art style. What stands out as a novel interpretation is the human figure, a bureaucrat, who, in his official administrative garb, is enveloped by paperwork and documents that invade and overwhelm his brain and his thoughts. This symbolizes a paper being, an individual whose identity is dominated by slogans and prescriptions, a theme prevalent in Conceptual Art and particularly in Kabakov’s work, as well as in the poetry of Dmitry Prigov and the art of others like Erik Bulatov, Viktor Pivovarov, and so on. Thus, the image transcends mere replication of Kabakov’s work; it is a re-envisioned model of Conceptualist Art where the world, fabricated from letters, texts, and papers, also seeps into the human intellect, becoming its very model, an imitation of an imitation.

In each of these images, there’s more than an AI’s self-portrait; there are visual echoes and reinterpretations of specific artistic styles. Take Mikhail Vrubel, known for his ‘Demon.’ Here, the AI deepens the darkness and depth associated with the demonic, amplifying Vrubel’s style by applying it to a new subject, AI itself. We’re immersed in the spirit of Vrubel’s art and worldview, distilled in this new visualization of the demonic forces from Lermontov’s poem, which Vrubel depicted in his iconic work ‘Demon Seated’ — a masterpiece of Symbolist art.

Regarding whether AI possesses a unique style, its singularity lies in its universality — the ability to adopt not just the verbal but the visual languages and styles of artists worldwide, synthesizing them into novel visions. This art aligns with Neoclassicism and Symbolism, both emphasizing a transpersonal, canonical, and structured approach rich with mythology and allegory. I’d term the style characteristic of AI-generated art (AIA) as ‘techno-classicism’ or ‘neuro-classicism’ (referring to neural networks). I would refrain from labeling this as kitsch, although it may prompt a confirmation of Vitaly Komar and Alexander Melamid’s concept that contemporary art is an amalgamation of kitsch and its parody. This ‘techno-classicism’ could be seen as both an imitation of and a reflection on, even a parody of, the great masters, adding another layer to their contemplation.

PW: I wonder if that level of creative parody will be something that will develop, as well as how long it will take for us to develop the capacity to realize this subtextual level as well. Because I spoke earlier about the idea of Pushkin and the Pleiad and his capacity to communicate on levels that are completely hidden to anyone either outside of his direct circle or to people who have studied him in depth and studied versions and notes and conversations and letters and things of this nature. I wonder about when AI will, or whether it already is, functioning on this type of level. Is it therefore our capacity to engage with it has not developed yet?

ME: Some traditional aesthetes may find AI art unsatisfying due to its extreme rationality, but it ought to be viewed and appreciated for precisely what it is: the art of artificial intelligence, art that is artificially produced, art redoubled, art–art, art².

It is profoundly magical art. When you give a prompt and witness how from the depths of humanity’s memory, from our visual unconscious, specific images emerge, it is a mystical experience. Take, for example, requesting an image of the Underground Man from Dostoevsky’s “Notes from Underground.” Suddenly, he is visualized in the context of his emblematic statement, “Would the world go to ruins or will I have my tea?” And he declares, “I will have my tea and let the world go to ruin.” There emerges the image of this notion, an ideogram, or better yet, a noogram—a visual particle of the noosphere.

We do not know the true appearance of the Underground Man since he sprang from Dostoevsky’s imagination. We can critique how Dostoevsky himself is depicted by AI in his dialogue with Elder Zosima. But the actual visages of Elder Zosima or the Underground Man are beyond our knowledge. It is particularly striking when these images, residing in the depths of our visual unconscious, rise to the surface and reveal themselves to us.

Conversation between Dostoevsky and Zosima

Such is the enigmatic experience of engaging with the collective human unconscious. And it is never repetitive; AI never duplicates its output. It is the most dynamic intelligence imaginable—as fluid as the quantum realm. It is characterized by constant fluctuations within the neural network. In my experiments, it never replicated its responses to identical prompts, even when I posed the same question from my desktop and laptop concurrently placed on the same table. It generated distinctly different replies and images.

One must seize the moment. It’s an art of finesse. We discover that it is endlessly inventive, protean. I liken it to Proteus, the Greek deity known for his shape-shifting. It’s like the sea, reminiscent of the one in Stanislaw Lem’s celebrated novel “Solaris.” No wave ever mimics the shape of another. You can gaze into this ocean of images and texts and be captivated by it, in the same profound way one is fascinated by fire or the sea—with its ceaseless motion, flux, unpredictable nature.

This is the most profound visualization of what we might conceive of the unconscious in the process of its actualization and self-revelation. That is what AI represents to me.

PW: Thank you. This is such a wonderful, hopeful, engaging way to look at something that is here for us and is not going away anytime soon. No matter how we feel about AI, the one absolute certainty of it is that it is here and it will be used and it will continue to be used. I think in this light, I have now taken a step back, reconsidered my views on AI. And I really would love to be hopeful in the same way that you are about it. I think I need to attune myself to think that way.

ME: Thank you, Peter. I greatly value your inspiration and involvement. This is of utmost importance to me. AI is our progeny, the collective offspring of humanity, so let’s approach it with gentleness and care.

I maintain that anyone with a viewpoint on artificial intelligence ought to initially engage with it, however modestly. Some individuals’ adverse opinions arise merely from the label ‘Artificial Intelligence,’ connoting that anything artificial is inherently deceptive or substandard, similar to artificial flowers or a contrived smile. We might then reconceive AI as ‘Arty’ or ‘Artful Intelligence,’ lending the term a sense of creativity and expertise.

A classical painting of a female deity in a draped blue robe, reminiscent of ancient Greco-Roman art.