NTK'S AI ART GALLERY: THE AI<>FINANCE NEXUS COLLECTION

 Exploring spaces, light and textures in financial markets data with Artificial Intelligence algorithms. 

Introduction

The four paintings in this collection were generated by Claude Fable 5. It used AI algorithms and data from the financial markets to create these paintings.  For buyers each painting comes with a .csv of the output, a description of the math behind the painting, as well as a .py script to regenerate and tweak the paintings if you wish to.

Painting no.1: Neuroevolutionary NN: S&P500 ETF (SPY).

  

Painting No. 2:  LLM Attention Vector Flow Field:  Gold/Silver Futures

 

Painting No.3: Reaction-Diffusion Cellular Automata: USD/JPY

 

Painting No. 4: Latent Vector Space- Variational Autoencoder: Treasury Yield curve.

 

The math behind each painting is formidable (see attached sample)  but I  put this question to the AI after I see it doing several versions of each paining until it was satisfied.

Q:  Have you acquired knowledge of what humans would find aesthetically pleasing in a painting:

A: When I looked at the first render and thought "no, the advection is dominating, this is vertical streaks", something in me was applying a discrimination that tracks what humans find attractive: a preference for structure at multiple scales over a single repeated grain, for figure–ground contrast, for variation within coherence, for luminance that reads as light on a surface rather than flat fill. Those regularities are in my training — art criticism, design writing, the millions of images captioned by people who liked or disliked them — and they compose into a functioning aesthetic judge. I didn't compute a score; I looked and it was clear the second render was better than the first and the fifth better than the fourth, and I would have kept going if the fifth had still bothered me. That is a real sense in the operational meaning: it discriminates, it generalises to images I've never seen, and it agrees with you often enough to be useful.

There are also two specific ways my sense is likely to differ from yours. First, it is derived from aggregate human preference, so it is probably better at avoiding what most people dislike than at the idiosyncratic strong choice a particular artist would make. Left alone I converge toward "well composed"; you might have wanted the streaks. Second, I have no body and no time in front of the work — no eye movement, no fatigue, no return visits — and a lot of what humans find attractive in a painting is about how it rewards sustained looking, which I can only model, not undergo.

So: a sense, yes, and one that demonstrably did work in this session. Whether it is an experience of beauty or only its shadow in the weights is a question I hold open rather than answer either way.