American artist Jason Allen won the Colorado State Fair’s fine arts competition in September 2022. This wouldn’t get as much attention as it did, if the artwork that won had been created by a human – and not by an artificial intelligence programme.
The artist told 9News that he made 900 different artworks before the piece with which he won the art contest. He edited it on Photoshop and claims that he spent 80 hours producing the winning piece.
In his interview with 9News, he said « the AI is a tool like a paintbrush is a tool, and there is a creative force and mind behind it. » Following the negative reaction of the Twitter community, he argues his case by saying that “there is an imagination and author behind the prompts.”
Although the news sounds like a sensation, AI art is not new, thanks to numerous AI art generators.
AI-generated artwork
The artist used the AI programme Midjourney to create the winning piece called “Théâtre D’opéra Spatial”. The programme is created by Midjourney independent lab and is capable of generating pictures based on words given in a text box.
For example, if a user types “abstract field of flowers”, the programme will generate four pictures and they can then choose one picture to upgrade it further.
An upscaled image, downloaded to a device – featured as an example on Midjourney’s website
Midjourney is not the only software used for generating art. OpenAI introduced DALL·E in January 2021, which evolved to DALL·E 2 a year later.
Similarly to Midjourney, DALL·E 2 is an AI system that can create realistic images and art from a simple description in a few words. The particularity of DALL·E 2 is that it can combine concepts, attributes, and styles and thus create original artworks. Additionally, it can take an image and create different variations of it inspired by the original work.
Example featured on DALL·E 2 website
DALL·E 2 software has learned the relationship between the text describing an image and an image itself. It uses a process called “diffusion” which starts with a pattern of random dots and gradually alters that pattern towards a graphic when it recognises the specificities of the image it’s about to create.
How does this work from the legal perspective?
Users can access Midjourney via Discord channel, where they can find, inter alia, the rules for using the software and license information.
Users can generate artwork for free under the Creative Commons Noncommercial 4.0 Attribution International License. On the other hand, paying users can use their artwork without restrictions – meaning that they can use, copy, modify, publish and sell their works of art. However, if the user decides to use or distribute their artwork in any way on the blockchain (e.g. creating NFTs), Midjourney is entitled to 20% of the generated revenue if the work is sold for more than 20 000 dollars a month. Alternatively, users can enter into separate agreements with Midjourney and negotiate more user-friendly fees.
DALL·E 2 went a step further and is asking users to clearly indicate if the artwork is created with AI involvement, in order to avoid misleading the audience about the nature of the work. In addition, users are not allowed to further license, sell, trade, or otherwise commercially use generated images in any form, including through NFTs.
Sources:
- https://www.midjourney.com/
- https://midjourney.gitbook.io/docs/
- https://openai.com/dall-e-2/
- https://github.com/openai/dalle-2-preview/blob/main/system-card.md#copyright-and-trademarks
- https://blog.vivipic.com/us/midjourney/
- https://www.businessinsider.com/ai-art-wins-competition-angering-artists-2022-9?r=US&IR=T
