Can AI Create Art That Truly Moves You? An Open Research Problem

Can AI create art that stirs a real aesthetic emotion in people, and is genuinely its own, not a copy of a style? I recently read Clive Bell's "Art" (1914). Bell argues that every real work of visual art shares one quality, which he calls "significant form": lines, colours and shapes arranged in a way that moves us. He also argues that some people feel it and others can't, that it takes you "out of life into ecstasy", and that only artists who feel it from the inside can create it. My intuition is that AI will not be able to do this. People can't explain what stirs them, so it isn't written down anywhere, and AI learns mostly from text. But these models keep getting better, and nobody knows the answer yet. So I want to test it properly: 1. Give a model the book. 2. Ask it to make genuinely new work, not imitations. 3. Ask a blind panel of people who truly feel aesthetic emotion to judge it, alongside real masterpieces. If it works, art can be decomposed into a program. If it fails, that tells us something too. WORK ON THIS WITH ME: This is a 5–6 month research project with a team of 3–4 people. I'm looking for: • An artist or designer who knows the feeling from the inside • An art historian or critic who knows Bell, Fry and formalism • An ML researcher in generative models or reward learning • A psychologist or neuroscientist who studies aesthetic experience Email me at [email protected]. Tell me who you are, which question in the video pulls you most, and one work of art that moves you, your portfoilo and relevant work. Read the book free: https://www.gutenberg.org/ebooks/16917 Images: book plates from Clive Bell, "Art" (1914), University of Toronto copy via the Internet Archive. Paintings by Cézanne, Giotto, Piero della Francesca, W. P. Frith and Luke Fildes, Chartres Cathedral and the San Vitale mosaics are public domain, via Wikimedia Commons. Hagia Sophia interior: photo by Ronan Reinart, CC BY-SA 3.0. Statue of Gudea (Louvre): photo by Darafsh, CC BY-SA 3.0.