-latent (2026)

Things that can be calculated and things that are vividly experienced by the body.

——What's in between?

With the advent of generative AI and large-scale language models, the idea that the world can be described and generated by language and mathematical expressions has become more realistic. However, between drinking water when you're thirsty or feeling the wind in the shade of a tree, I think there's something that spills down from there no matter how you think about it.

This painting is a reconstructed image after removing features extracted by AI from my profile image. Does meaning emerge even from fragments discarded by AI as errors or noise?

As a new work in the LATENT series, which explores human cognition and senses, I used myself as a motif for the first time. Incidentally, the composition and size of this work refer to a famous self-portrait. Do you understand?

Profile photos are compressed into latent space (Latent Space) with a generative AI variational autoencoder (VAE), and differences from the original image are extracted (Significant 1x). AI determined that it was “unimportant,” emphasized information that was not held as noise using ternalization processing (Tri-tone), and reconstructed by hand onto the canvas based on that structure.

The composition and size of the work refer to Van Gogh's Self-Portrait (*).

* https://www.musee-orsay.fr/en/artworks/portrait-de-lartiste-747