Futuro Fortissimo Archive
🎼 ff.30 DALL-E: it makes art
Artificial intelligence that paints anything you want
Published · Michele Merelli
- Can it help us understand Picasso?
- Will it show the Fisher-Price people their next children's toy (a crucifix)?
- How fast is this technology improving?
👨🏫 ff.30.1 A short introduction
Substack ↗Before diving into DALL-E 2's art, the explainer.
DALL-E is a name that blends the art of Salvador Dalí and Disney's cute little robot. It's an artificial intelligence developed by the OpenAI folks (the same ones who beat the world GO champion; we talked about it in 🎼 ff.18 Between digital art and AI) able to generate images consistent with a text input.
DALL-E can also place, in a given image, any kind of object we want, like a dog in a museum. And it does so by adapting the style to the location:
For now the model is accessible only to a few people, given the ethical issues that could arise from its misuse.
😂 ff.30.2 A few funny examples
Substack ↗Let's see it at work, with the related descriptions.
“An astronaut lounging in a tropical resort in space, photorealistic style”.
“An astronaut lounging in a tropical resort in space, pixel art style”.
“Variations of Vermeer's Girl with a Pearl Earring”.
😱 ff.30.3 A few less funny examples
Substack ↗A collective (called Hugging Face) recreated the model and opened it to the public. You can play with it, although the resolution is lower, given the hardware limits for training the AI.
The Twitter account Weirddalle uses this mini-model to generate images somewhere between distressing and hilarious:
“Rusty Teletubby-shaped wreck on the ocean floor”.
“A urinal developed by Razer Gaming”.
“SpaceX launching a crucifix into space”.
“Fisher-Price-style crucifixes”.
👂 ff.30.4 Understanding Picasso
Substack ↗Another Twitter user suggested using the software's regression ability (when it's asked, for example, to show the dynamic steps from one image to another) to analyse Picasso's sketches.
Starting and ending on the artist's drawings, what's shown in between is a wandering / imagining of intermediate phases between the two forms. Thanks to DALL-E, then, we see the creative mental process that might have led Picasso from one idea to the next. The simplifications and “prototypes” between the various sketches.
Sarebbe molto interessante preservare, a livello visivo, non lefasi ma le metamorfosi di un immagine. In questo modo, potremmo scoprire il percorso seguito dal cervello per materializzare un sogno. Pablo Picasso,1935
📈 ff.30.5 The improvements in just one year
Substack ↗What's phenomenal about artificial intelligence is the speed of its development. For years people have predicted that the physical (quantum) limits of Moore's Law would halt progress in computing power. The reality is that humans keep inventing loads of ways around them.
In one year, DALL-E (now version 2.0) improved its results, reaching 4 times the resolution. Seeing is believing.
“An avocado-shaped chair”.
DALL-E 1
DALL-E 2
The next step? Probably video creation. Imagine the power of a tool that can depict, for example, any film plot that comes to mind. And maybe do it in an afternoon.
The creator economy of the future will put today's influencers to shame.
👋 And that's all for this fortissimo week! 🔜 In the next ff we'll talk about how to fill the summer and your time under the beach umbrella. See you soon, Michele
Want to support this project? Subscribe to the newsletter and share it with a friend on WhatsApp
👇🔥Subscribe
Or buy me a coffee 😏
☕
And remember to leave a little heart below.
❤️