DreamUp AI Art Generator Illustrates the Collision Between Artistic Consent and AI Datasets

I entered the prompts that produced these images.

DreamUp

Behind the AI ​​art is data – countless images used to train the AI ​​art generator. When a user enters a prompt, the generator splits these existing images to produce a new combination of colors and shapes, arguably a new work of art.

AI art generators draw information from what they are exposed to, just like human artists do. But there is a difference in scale in how humans go through input, process it and try to produce something new compared to AI.

When DeviantArt released its AI art generator, DreamUp, there was an immediate reaction from its community that the community’s works were automatically selected for use in AI datasets.

Creators will need to manually opt out to protect their work from future AI image training. However, the opt-out request will only take effect once their work has likely been used to train DreamUp in some capacity.

Also: How to use DALL•E 2 to turn your wildest imaginations into tangible art

DreamUp is an AI art generator that creates art based on prompts, with roots in Stable Diffusion. His ability to create art is based on content taken from the web without the notice or permission of the artists who made the work. An artist on Twitter summed up the process well:

Content pulled from DeviantArt and other sources goes into a LAION dataset, which is then processed through machine learning, ultimately resulting in a 3D model. The opt-out request occurs before any machine learning, but does not apply to initial learning.

DeviantArt responded to their concerns in an update, saying that:

  • They will not use artwork provided to the DeviantArt community in DreamUp or other AI models or training kits.
  • They did not consent to images being deleted from the site by third parties.
  • To stop future unauthorized use of art when training AI models, they are releasing a “noai” flag that will notify AI models that the artist does not want their work to be used. This won’t guarantee that the artist’s wishes will be respected, but it’s a start.

I tested DreamUp after seeing that DeviantArt had addressed the issues raised by artists. I had to register to use it and had five free prompts to use. In the quick directions it reads: “Images inspired by other artists must clearly mention this artist when published.

Also: The people who build AI are the ones who need AI the most

My image was inspired by a dream I had (thanks, subconscious) and my prompt was “A shoe-shaped car driving down a highway.” These are the three images that DreamUp gave me in less than a minute:

DreamUp images of cars

DreamUp

Hmmm, I was imagining a different shape, but I don’t mind the bottom pic.

Image of a DreamUp car

DreamUp

I don’t feel comfortable taking credit for these images as I don’t think I had enough of a hand in creating them, so I’ll give credit to the generator itself.

I also tried another prompt, “couple driving a rocket ship through space”, and got the following three new images:

DreamUp couple in rocket ship images

DreamUp

My favorite of the three was the comic:

Images of DreamUp rockets

DreamUp

I finally tried a prompt with my little brother’s favorite things “galactic dragon stephen curry”. His favorite of the three was this one, even though Steph Curry’s number is 30 and I’m sure he doesn’t play for the Goory Boxty Terrors?

Image by DreamUp Steph Curry

DreamUp

These images were not cohesive and had a melting, blended quality. Still, it was fun experimenting with DreamUp. I see it more as a point of inspiration for art than as a replacement.

The ease with which you can input something into a generator and create an image has led to art communities full of AI-generated images.

Some communities have banned AI art entirely. This plays into a wider debate about what gives art value and how AI fits into the puzzle.

Also: The real goal of AI may no longer be intelligence

Should it be a problem that countless artwork data points are being used to train an AI that spits out art in less than a minute, art that wins competitions no less?

The sheer amount of art you can create with AI dwarfs in volume the only artwork created by humans that requires hours of thought and labor. I think this could lead to several outcomes:

  1. AI art is used more for inspiration than as an end result in the human creative process.
  2. Art enthusiasts prioritizing “handcrafted” works by humans over AI-generated, more assembly-line parts.
  3. Other creative industries, such as music and writing, eventually developed their own versions of generators.

The potential with AI art generators is huge, and so are the possible consequences.

Leave a Comment

Your email address will not be published. Required fields are marked *