Thanks Mason!

Crossover operations (combining elements from different designs) substantially boost novelty. The algorithm performs crossover in the semantic space of language, enabling conceptual combinations. We can also measure the expanding coverage of the latent space.

Ask ChatGPT to "design a creative time display" and you'll get pretty similar radial designs. AI tends toward average outputs. Lluminate counters this by systematically exploring creative possibility spaces.

New research project: Lluminate - an evolutionary algorithm that helps LLMs break free from generating predictable, similar outputs. Combining evolutionary principles with creative thinking strategies can illuminate the space of possibilities.
www.joelsimon.net/lluminate

It's exploring two latent spaces since mixing happens with image embeds (flux ip adapter with uniform crossover) and the sliders are found directions in clip-text space 🎚️

This feels like a culmination of six years of latent space creative interfaces (👴), but also a return to the joy of instinctual, unintentional wandering—like the original GanBreeder—rather than being confined by the limits of language and the burdens of articulating intent with text-to-X.

I'm releasing a new latent space explorer 🧭🎨 — a creative tool for generating and evolving images.

It's tree-based and designed for creative flow and open-ended discovery, using image mixing, branching, and editing.

You can try it out now!
www.artbreeder.com/tools/splicer2

😄✋🌐

Testing generative video with Puppets.app. Many ideas cannot be captured by language, and hand-controlled puppet rigs offer a playful sense of craft.

We need new metaphors and interfaces for video. Shadow puppets are simple yet timeless.

nice!

😂

Thank you! :)

WU-TANG IS FOREVER

THE SAGA CONTINUES

Inspired by a visit to the incredible cave paintings at Lascaux last week I made a site where you can make your own hand paintings 🖐️

www.joelsimon.net/forgotten-dr...