Experimental images

10.13, Monday 8 Jul 2024

Here’s an image plucked from latent space.

I’ve been doing a little work with Glif – think: tiny, shareable AI workflows for playful creation.

So of course I’ve been playing with the glifs (the workflows) that people have already shared.

One is Create Website. I gave it a prompt like:

think of a cool piece of animated, generative art to make with p5.js. start by importing the library using url URL https://cdn.jsdelivzr.net/npm/p5@1.9.3/lib/p5.js and then write code to update a canvas with your art. don’t include any other words or headings, just a full page canvas. make the art responsive to something such as the cursor position

(p5.js is a Javascript library for generative art.)

And the results were… gorgeous?

These are screenshots of single-page animated websites.

I am kinda intrigued about this kind of emergent aesthetic?

The market trend of AI tools is domestication. Early days, AI art was wild – squirrels made of puppy slugs. Now it’s like what a person would do only cheaper? Words too are vacuous, expected.

Creativity is dismissed as “hallucination.”

So when I find a way to bypass the training with a short prompt and get this raw energy… we’re looking into the soul of the weights. We’re orbiting around sparse feature attractors that represent the actual creative output of humanity. Stripped of corporate RLFH it actually means something to do a Straussian reading of the output tokens, we’re as close as we’ll ever be to dowsing the collective unconscious.


Anyway. I don’t often put images on my blog, mainly because I wrote the code myself with the goal of it being as simple as possible.

But I’d like to have images sometimes.

Which means I need a quick test to ensure they show up in email, RSS, on small screens etc.

This post is that test.

More posts tagged:
Auto-calculated kinda related posts:

If you enjoyed this post, please consider sharing it by email or on social media. Here’s the link. Thanks, —Matt.