Not knowing how to draw will cease to be an impediment to being able to make graphic reality what one imagines, or at least what can be described in words. All thanks to Dalle-2, an Artificial Intelligence capable of generate illustrations with high-quality graphics from descriptions they don’t even need to be particularly detailed or specific.
Dalle 2 will be very useful for people with little ability to draw
The investor and disseminator Javi López has explained in a Twitter thread some of the characteristics and advances of Dalle-2, giving as an example the result obtained after describing something like “an astronaut raccoon with the cosmos reflected in the glass of his diving suit dreams of the stars”.
Although it may not seem like it, behind this illustration there is not the imagination or the talent of an artist, but the work and “sensitivity” of an artificial intelligence.
Dalle-2 received only this text: “A raccoon astronaut with the cosmos reflecting on the glass of his helmet dreaming of the stars” and generated this image. Flipa. pic.twitter.com/oe7iXil4Ct
– Javi López ⛩️ (@javilop) April 8, 2022
It is not yet fully open to the public, but researchers, developers or people interested in artificial intelligence can request access to the project from the Dalle-2 website. Some are already sharing the results obtained on the Internet, such as the illustrator Carmen Lau, who has shared a drawing made by herself and the result obtained by Dalle-2 after describing the same illustration in words. Lau herself acknowledges that he likes more the work created by Dalle-2.
I tried using DALL-E to recreate one of my own digital illustrations by inputting a verbal description of it. Left is my original, painted in Photoshop with a Wacom tablet. Right is the generated one. I actually think I like it more, esp. colors! But it does worse on fine details pic.twitter.com/JvbxDwCn9Z
– Carmen Lau (@carmenleelau) April 8, 2022
Lau explains that in his experience Dalle-2 works better if it does not receive extremely long or complex instructions and that it can be very helpful for people with little or no artistic or creative abilityas well as to get preliminary models of artistic concepts.
At the moment, Dalle-2 is still in the early stages of its development and, like other artificial intelligences specialized in specific tasks, years of work, evolution and contributions from diverse teams are missing before it can perfect its capabilities, but the truth is that even now the examples shown of what Dalle-2 is capable of achieving, let’s remember, based on few-word descriptions, are surprising and some of them really beautiful.
Not only that, but also, this artificial intelligence is capable of modifying photographs by introducing changes according to the assigned instructions. This way you can replace, for example, a dog with a cat in an image. As could easily be done using Photoshop but made even simpler: just tell Dalle-2 to “change the dog to a cute cat”.
Dalle-2 allows today to retouch images to the level of changing any element that is in them.
To the extreme of changing a dog for a cat and that it is perfectly integrated. pic.twitter.com/UTrRfjFZIP
– Javi López ⛩️ (@javilop) April 12, 2022
Multiple questions arise regarding the results obtained with Dalle-2, even in its still preliminary phases: Are illustrators’ jobs in danger? Will anyone be able to become an “illustrator” by dictating the proper instructions to Dalle-2?
Can we believe our eyes in front of a photograph? Will this type of artificial intelligence be integrated into the mobiles themselves to modify/retouch photographs through voice instructions? What about intellectual property rights? Is Dalle-2 the “creator” of these illustrations or just a tool and credit goes to whoever dictates the description of what the image should contain?
Will artificial intelligence end up replacing artistic creation? Are there already some capable of creating possessed imitating the great authors? Will androids dream of electric sheep?