#87 - AIs that generate images

🏞️ AI: we have reached a milestone!

If there is one area that has fascinated me in recent months, it is the generation of images based on artificial intelligence models.

In 2021, Dall-E gives us a first preview by generating images from text queries. It's nice but could be improved.

In April 2002, Dall-E 2 gave us a slap in the face by creating images and works of art by combining complex queries and graphic styles. It amazed us so much that with Rémi we recorded a podcast on the subject (see the video below 👇).

Problem in that very few people had access to it.

At the end of August we discovered stable diffusion which offers similar functionalities with the provision of APIs to integrate this into applications.

And then we realize that we are experiencing a tipping point in graphic creation.

Because it is now a concrete reality: we generates images that look real as easily as a Google query.

Some examples below with a new feature of the Photoroom application (which uses stable diffusion) accessible without authentication:

You can imagine lots of use cases and even if it raises a lot of questions:

• Who owns these images generated from other images created on the internet
• Can we have a commercial use of them
• What about certain activities like photographers and illustrators?
• Is this the end of image banks?

In fact it's quite amusing, I barely had time to ask myself the question that a startup had launched the first image bank generated from AI:

Well in any case it's just incredible and you have to try it!

Okay and to finish on the generation of images I wanted to share with you this thread from Greg Isenberg which shows what certain deceased celebrities would look like if they were still alive.

Troubling.

Have a great week and see you next week to talk about brains and organization 🤯.