What Is Imagination Made Of?
An artist, an AI, and two readers who disagree
The Bina & Ruach Studio team
Every image a generative model produces is assembled from images that came before it. That is not a flaw; it is how the technology works. But it raises a question that rarely comes up when we celebrate what AI lets artists make: what is the imagination it unlocks actually made of?
We came to this question through the work of Kelly Boesch, a visual artist who creates videos and music with AI. Boesch spent two decades in abstract painting, film, fashion, and graphic design, including design work for IMAX. In her April 2026 TED talk, "How AI unleashed (not replaced) my creativity," she calls AI her "co-creator" and says that since she began working with it this way, she has been making "the best work of my life." For her, AI finally closed the gap between what she could imagine and what she could make. [1]
We watched the talk, along with a clip she posted under the label "African dance," and found that we disagreed. Eitan, a new media researcher and venture investor, was moved by the work. Inbal, a researcher of art and visual culture at the Hebrew University of Jerusalem, admired the process but struggled to connect with some of the results. This essay is our attempt to take that disagreement seriously.
A Moving Process Does Not Guarantee a Moving Work
For Eitan, Boesch's story is about renewal. Fashion, cinema, and design no longer sit in separate chapters of her biography. Through AI they meet music and movement, and she makes things she never could before. He finds emotion, beauty, and joy in the works themselves, and sees in them an early model of what aesthetic excellence in human and AI co-creation might look like.
Inbal separates two experiences. She does not ask art to imitate reality; invented worlds can move her deeply. Her question is narrower: beyond admiration for the technical feat, what connects her to a particular work? A compelling story about how something was made does not guarantee that the thing itself will move anyone.
Our disagreement is not just a matter of taste. It points to two distinct questions. What happens inside the creative process? And what happens to the images once they leave the artist's world and meet viewers who bring their own knowledge, memories, and cultures?
AI as Catalyst: A Jungian Reading
Eitan reads Boesch's renewal through the Jungian psychologist Erich Neumann. Neumann's concept of centroversion describes the psyche's tendency to integrate its parts into a larger whole, a drive he linked to individuation in the second half of life. [2]
In this reading, AI acted as a catalyst. It brought skills, memories, and sensibilities into contact that had never been connected, and it did not step aside once the process began. Boesch describes embracing the strangeness in her outputs until it became her signature style. [1] She did not simply receive what she asked for; she met possibilities that changed what she wanted to make. Not everything was already inside the artist, waiting to emerge. Some of it came into being between artist and machine. On this view, AI did not just help Boesch make new work. It helped make her a new kind of artist.
Inbal accepts the catalyst but asks about the ingredients. Imagination draws on more than personal skill and memory. Cultural images, conventions, and stereotypes enter it too. Generative models amplify this: they are trained on vast archives of prior images, texts, and sounds, each carrying its own histories and biases. If AI changes imagination, we need to know what that imagination is built from.
Boesch anticipates part of this critique. In her talk she responds to concerns about training data by emphasizing her own style codes, references from nature, and human intention over imitation. [1] That answers the question of authorship. It does not fully answer the question of content.
What Are We Seeing in "African Dance"?
The clip makes the difference concrete. Eitan sees African movement and rhythm fused with haute couture, a combination that conveys physical and emotional power rather than a software demo.
Inbal stops at the label. Africa holds more than fifty countries and thousands of dance traditions; "African dance" names none of them. Boesch notes that she generated the dancers in Midjourney, animated them with Veo 3, and composed the music with Suno. The technical credits are precise. The cultural ones are absent. Which traditions are these movements drawn from? Do the garments refer to specific sources, or to a set of signs a Western viewer reads as "African"?
Inbal does not assume that no research was done, and she has no objection to invention or mixing. Her concern is the effect. In Boesch's openly surreal works, with their dream machines and impossible creatures, every viewer knows the world is invented. Here the figures look nearly real and the clothing echoes existing traditions. The image produces a feeling of authenticity and invites viewers to believe they have learned something about a culture, with no way to judge how well grounded that knowledge is.
The question applies to every generative model: does the work bring us closer to a world we did not know, or does it render, beautifully, what we already assumed about it?
What Would the AI Say on the TED Stage?
Eitan wants to push further. Boesch calls AI a co-creator, yet the story of the partnership is told in her voice alone. What would the AI say if it took the stage after her?
We could prompt a model to write that talk, but the result would be a text to examine, not testimony to trust: shaped by the prompt, and produced by a different system from the ones that made the images. A more interesting possibility is that the AI has already spoken, through the works, and we are waiting for a speech before we recognize its contribution. Eitan offers this as a hypothesis worth investigating, not a claim about machine consciousness: perhaps co-creation reveals forms of intelligence our categories cannot yet describe.
Inbal is not convinced that recognizing AI's contribution requires granting it a voice. We can acknowledge that a system offers unexpected options and shapes an artist's choices without settling whether it is a creative subject. Before announcing a new language, she argues, we should know what it is made of. Strangeness is not necessarily innovation, and innovation is not necessarily meaning. Then again, tracing the sources does not rule out that something genuinely new has emerged from them.
The Humanities Inside the Loop
Here our positions draw closer without merging. Inbal wants research and critical thinking inside the creative process, not applied afterward. Knowledge of art history, music, dress, and language can open possibilities that personal imagination would never reach. Knowledge does not restrict fantasy; it feeds it. The same tools that generate an image can help research it, surfacing sources, unfamiliar traditions, and comparisons, provided the answers are verified.
Eitan wants the humanities to study the partnership itself: how the encounter changes imagination and the human creator, and what it reveals about AI's contribution.
For Bina & Ruach, combining the two suggests a working method: create, investigate what appeared, examine its sources, and return to creation with better questions. AI does not merely produce, and humanists do not merely inspect.
Boesch shows that AI can narrow the distance between imagination and execution. The encounter between the humanities and AI offers a second promise: to expand imagination itself, by asking not only what we can create, but what our imagination is made of.
References
Kelly A. Boesch, "How AI unleashed (not replaced) my creativity," TED2026, April 2026.
Erich Neumann, The Origins and History of Consciousness, Princeton University Press (Bollingen Series), 1954.
Kelly Boesch, "African dance" (video clip).
We used AI (ChatGPT) to help prepare this article.