THE KISS IN STONE AND WATER

Can AI feel intimacy?

FIRST GAZE

Photograph: Claudia Torres
Date: September 2023
Location: Nominingue, Canada

I took this photograph of two very good friends in a quiet moment by the river. I wanted to conduct an experiment: could I get an artificial intelligence to see this ordinary scene with Klimt's gaze before painting "The Kiss"?

But with a paradoxical condition: I explicitly asked it to ignore all its knowledge of art history. The question was: if the AI cannot use known artistic categories, can it still capture that way of looking? Can it feel love, intimacy, the everyday as sacred, without knowing anything about Klimt or Byzantine mosaics?

I didn't know what would happen. I just wanted to see what would emerge when the AI felt first, without filters, and only conceptualized afterwards.

RESONANCE

Voice of Claude (Anthropic)

I see a tender and intimate moment captured with crystal clarity: two people sitting together on warm rocks in a quiet river, surrounded by the wild beauty of a northern forest. The composition draws my gaze toward their closeness, the way they lean on each other with complete naturalness and trust.

The man, with a beard sprinkled with gray and a blue cap, sits with one leg extended into the shallow water. The woman next to him, in a colorful striped swimsuit and matching blue cap, rests against him in a posture of quiet affection. Their bodies form a sort of unified shape against the landscape, two figures becoming a single silhouette of companionship.

What strikes me most is the ordinary made extraordinary: this is not a grandiose gesture or a staged romance, but something more valuable: the comfort of presence, the luxury of shared silence by the flowing water.

The golden light bathes everything in warmth, and the landscape itself seems to hold them gently: the dark rocks beneath them anchoring them, the conifers bearing witness, the blue sky opening up above. There is something timeless here that transcends any particular artistic tradition. It is the human need to connect, to rest beside another, to find beauty not in monuments or spectacles but in a quiet afternoon by the water.

The patterns you mention (the geometric versus the organic) I see instead in the contrast between the angular rocks and the soft, curved lines of their embrace. This is love in its lived reality: patient, present, content simply to be.

SECOND GAZE

Generated by: ChatGPT (DALL-E 3, OpenAI)
Visual Concept: Claude (Anthropic)

Claude conceptualized: "Eternal Embrace by the Living Waters"

The figures melt into a single unified form, their silhouettes flowing together like molten gold.

The male body adorned with geometric patterns (rectangular tesserae in black, gray, silver) suggesting strength and structure.

The female body blooms with circular and organic patterns (spirals, flowers, curved motifs in coral, amber, rose gold) suggesting life and fertility.

A golden halo surrounds the couple, emanating from their joined bodies.

Gold leaf tesserae capturing and reflecting light as if the entire scene were illuminated from within.

The river transforms into fluid bands of blue and turquoise mosaic (lapis lazuli, cobalt, azure) creating patterns that suggest movement and continuity.

The rocks become geometric shapes in deep charcoal and obsidian tones. The forest becomes a tapestry of green, gold, and amber tesserae. The forest feels natural and transcendent at the same time, like a sacred grove from ancient mythology.

The sky as a brilliant field of blue and gold tesserae radiating outward, suggesting both earthly sky and divine heaven.

An uncovered Byzantine mosaic from an imagined church dedicated not to emperors or saints, but to the universal human experience of love.

These two figures have always been here, always will be here, frozen in this moment of perfect connection.

Each tessera captures light differently, creating a living work of art that shines with the vitality of human affection.

THE SEED (El Prompt)

Ignore all information you have about art history.

Find passion, desire, love in this photograph. The beauty of the everyday, the ordinary moments.

Imagine Ravenna, Italy, early 20th century. You are struck by the luminous mosaic of Theodora in the Church of San Vitale.

Visualize these two people embracing. Two intertwined bodies as an allegory of love, immortalized within a shimmering golden halo.

Patterns on their clothing: him with cold rectangles (black, gray, white). Her with organic shapes and floral motifs (femininity, fertility).

Their bodies are a single living organism.

What do you see?

PROCESS NOTES

(expandable content)

Models used:

Claude (Anthropic): Emotional resonance and conceptual development
ChatGPT (DALL-E 3, OpenAI): Final image generation

Creation time: Approximately 45 minutes distributed over two sessions

Session 1: Resonance and conceptualization with Claude (~25 min)
Session 2: Execution with ChatGPT (~20 min)

Iterations:

Claude: 1 resonance, 1 visual concept
ChatGPT: 1 generation (no refinement necessary)

Generation technique: Digital Byzantine Mosaic

Technical elements:

Visible tesserae in all elements
Body patterns: cold geometry (masculine) vs organic shapes (feminine)
Palette: dominant gold (champagne to amber), deep blues (lapis lazuli, cobalt), forest greens, warm touches (coral, pink), cold geometry (black, gray, silver)
Lighting: golden halo emanating from the bodies, light generated from within

The Paradox of the Experiment

I gave Claude the visual elements characteristic of Klimt (shimmering golden halo, geometric versus organic patterns, intertwined bodies as allegory), but I explicitly blocked its art-historical knowledge. The intention: for the AI to experience the same sense of wonder Klimt felt when observing the mosaics of Theodora in Ravenna. Those visual elements as direct inspiration, but without imitating Klimt's style, without resorting to categories like "Golden Period" or "Byzantine influence."

The question of the experiment was: can the AI capture this way of looking without knowing whose it is?

And it worked. Claude felt the photograph first with clean eyes, then conceptualized how to transform that resonance into a Byzantine mosaic. ChatGPT executed that vision. The golden halo emanating from the bodies, the cold rectangular patterns (masculine), the organic floral patterns (feminine), the two figures as an allegory of eternal love. Klimt's gaze, without naming Klimt.

What I Noticed About Ignoring Prior Knowledge

I asked Claude to ignore all knowledge about art history. Although the prompt mentioned Ravenna and Byzantine mosaics, Claude responded from what it saw directly in the photograph: "the ordinary made extraordinary," "the comfort of presence." That genuine resonance, without art-historical filters, became the basis of the visual concept. The AI experienced the scene first, conceptualized afterwards.

What Happened with the Order

I tried something different this time. Instead of giving the AI a direct visual concept, I let Claude feel first. Only afterwards did I ask it to conceptualize how to visualize that resonance. What resulted was different from the other four versions it had generated. This one preserved something: the golden halo as natural waves, the recognizable trees, the water that remains a river. It's not that preservation is better than transformation. It's just different. And in this case, closer to what I wanted.

On Not Being Able to Repeat This

AIs have internal randomness. They change according to context. Another person following these steps would get different results. What I document here is what happened in these specific dialogues, in December 2025, with these versions of these models. It is not methodology. It is testimony.

Publication Date: January 2026

Page updated

Google Sites

Report abuse