Analogies, Strangeization: How AIs Tango with our Brain.

I asked an AI, ruDALLE-XL, to make images of a dream.

This AI is a fun one, it’s easy to disturb her (therefore to inspire her).

It’s not the only one: I asked DALLE 2 (OpenAI, Google’s text-to-image AI) to draw me “Slow Melancholy” and I got this snake shell with leaves. Not bad, good picture composition and light. It looks like a photo, and that’s it.

Now, look at these pictures. The first one looks like clouds over a lake (but on the right, they look like something else). The second one shows some “things”, in the morning on a green land. But what are they? The third one shows the silhouette of a woman. She’s probably on the beach, looking a… well… what?

Our intelligence, or our way of understanding reality, is based on ANALOGY. When we see something, our brain is trained to “link” it with something we already know.

“Oh, it looks like a cloud”.

Douglas Hofstadter wrote an entire book about this idea. He shows us a page full of the letter “A”. Some of them we’ve never seen, but we know they’re “A”s.

It becomes interesting when we see the strangeness produced by AIs.

Image 1: a red sphere emerges from a… flower?

Image 2: something (an animal?), jumps on the sand?

Image 3: some fabric, or maybe a creature under the sea?

We would like to know, but we can’t.

It’s a bit disturbing, or ugly. It’s beautiful, maybe, who knows?

Is it a leaf? An animal? Who’s that lady? What does she think?

Could we write a poem about one of these 6 pictures? Invent a painting from it?

One quality of a good picture today is to stop the gaze, to stop the viewer’s eye, wandering fastly in the world made of images.

AI-made images, some of them, have this quality. Our brain stops because it tries to understand. Is it a bird in the night? Watching what? Why is the moon wrapped in this shape?

Strangeization is a way for the artist to “add strange elements” to a piece of art, in the purpose to catch back an audience who has seen everything and is hard to get attention from.

Here, the process is: make plenty of images with a prompt (words that describe), and choose one.

What will you make?

—

Strangeization Tool & Eyebrow Criteria (or type this word in my blog’s search engine).

Thanks for reading!

—

Digit Lovers vs Artists

There’s a pattern in photography.

Two kinds of photographers. Those who take photos with their camera. And those who play for hours on their Mac, with apps, with digits, HDR, and I correct this and that.

I already wrote articles about those.

Take a car.

You can be the tech type. You want to open it and fix this and that. Or you want to drive it to the lake and breath fresh air.

Today I’m very busy making images with Artificial intelligence. I have an Artstation (here: https://quick_eyed_sky.artstation.com/ ) and a YouTube channel here: https://www.youtube.com/c/JPRobocat

Very soon, I noticed the same pattern.

To make images, you need a Google Colab, which is a Python program with a bunch of settings, like a machine, and you enter a “prompt”, which is a “phrase describing what you want to see”.

There are 2 camps.

• Most humans are obsessed and focused on SETTINGS, they study these for weeks, very deeply, line by line. They’re obsessed with numbers, and digits.
• The other camp is obsessed with the PROMPT, which is like poetry. Choice of words. The way you present them. Their weight. Etc.

It’s always the same pattern.

• Camp 1 is very serious. They are logical. They are focused on their numbers. They want DETAILS, crunchy pictures, they want lines. Techs.
• Camp 2 is very casual. They try things. They read poetry. They enter lyrics and try to find new artists. They want mood, light, and composition. Poets.

You know what? Both camps are OK. But I’m in the 2nd one. You knew that already, right?

Do you watch the beach and the mountain, or do you watch your car’s engine?

Thanks for reading!

Do you focus on the prompt or on the digits everywhere? What about “clamp_max”: 0.05? Should I double it? Or should I add “ominous sky” in the prompt? Where do you like to work?

“text_prompts”: {
“0”: [
“Greg Rutkowski, long butterfly airship in the summer sky, Artstation”
],
“100”: [
“This set of prompts starts at frame 100”,
“This prompt has weight five:5”
]
},
“image_prompts”: {},
“clip_guidance_scale”: 50000,
“tv_scale”: 0,
“range_scale”: 150,
“sat_scale”: 0,
“cutn_batches”: 4,
“max_frames”: 10000,
“interp_spline”: “Linear”,
“init_image”: null,
“init_scale”: 1000,
“skip_steps”: 0,
“frames_scale”: 1500,
“frames_skip_steps”: “60%”,
“perlin_init”: false,
“perlin_mode”: “mixed”,
“skip_augs”: false,
“randomize_class”: true,
“clip_denoised”: false,
“clamp_grad”: true,
“clamp_max”: 0.05,
“seed”: 2397292033,
“fuzzy_prompt”: false,
“rand_mag”: 0.05,
“eta”: 0.8,
“width”: 1600,
“height”: 832,
“diffusion_model”: “512x512_diffusion_uncond_finetune_008100”,
“use_secondary_model”: true,
“steps”: 350,
“diffusion_steps”: 700,
“ViTB32”: true,
“ViTB16”: true,
“ViTL14”: false,
“RN101”: false,
“RN50”: true,
“RN50x4”: false,
“RN50x16”: false,
“RN50x64”: false,
“cut_overview”: “[12]400+[4]600″,
“cut_innercut”: “[4]400+[12]600″,
“cut_ic_pow”: 1,
“cut_icgray_p”: “[0.2]400+[0]600″,
“key_frames”: true,
“angle”: “0:(0)”,
“zoom”: “0: (1), 10: (1.05)”,
“translation_x”: “0: (0)”,
“translation_y”: “0: (0)”,
“video_init_path”: “/content/training.mp4”,
“extract_nth_frame”: 2
}