Hello everyone! I hope you guys are doing well! π
I want to talk about something very controversial these days: AI art. AI art and its ethicality have been controversial since the medium began, but you will need some context to understand why it is this way. So here is a basic summary for anyone out of the loop about how AI art models work:
- 1.) An AI model is trained on images that the person or company making the model made or, more likely, scraped off the internet (Reddit, Instagram, DeviantArt, etc.)
- This is pretty much the same process that trains LLMs (Large Language Models). Think ChatGPT, Gemini, LLama, etc.
- 2.) After creating these models, they are released to the public through websites like Tensor.Art where people can create images BASED on the database of images on which the model was trained.
- Models that can fit into this category include Stable Diffusion, DALL-E, and MidJourney, which are all DMs (Diffusion Models).
The problem with these models is that the images/art they are trained on are usually used WITHOUT the explicit permission of the artist/person who initially made the image. Obviously, this is pretty sketchy and possibly against copyright law, which is why artists are quite mad when pieces of art that are made with AI come out and become viral.
SIDENOTE: OpenAI is actually being sued by the New York Times over copyright infringement since the company trained ChatGPT using NYT articles. The lawsuit is pending as of writing this blog post.
I will show you two examples, with both pieces being based on a VOCALOID, specifically Hatsune Miku, and both have been released on Twitter (now X). The first example was released on September 14th, 2024, by the X account @pon_pon_pon_ai:
This image of Miku (which had 22.5 million views as of writing this post) sparked backlash within the art and VOCALOID communities because it didn’t look very good and was made using AI. People then started making versions of this art to fight against AI art. Some examples are below:
This trend of redoing the original piece lasted for some time but died down a couple of weeks later. But this is not the last time this year that we will see an AI art piece using Hatsune Miku as the centerpiece. On November 13th, an X user going by @furkiwi released this artwork of Miky as a cat/furry:
When I saw this image for the first time, I had no idea that it was made with AI. It looks flawless when looking at it from a glance, but small details, such as her paws and especially her tail, gave it away to me that this was done using AI. As you might be able to guess, this also sparked backlash, with people redoing this piece of art as well. But, after about a week, they released the model they used to create the image. So, for fun, I gave it a shot, and as someone who doesn’t like to make art, this is unbelievable:
Model Link: https://tensor.art/template/794143984489425932
These images are unbelievably detailed, and except for some minor details, they don’t look like they were created with AI. But this raises the scary thought of artists being replaced by AI trained on their art, and I agree.
Here is my position on AI art after trying a model for the first time and hearing both sides of the argument: AI art is incredible and absolutely a fantastic breakthrough, but like all AI, it needs to be created and used responsibly. This means ensuring that artists get paid a decent amount of money for their art if it is used to train AI models and that people can tell if the art was made by AI or by a human. Artists will not be able to stop AI: the companies making these image models are WAY TOO invested in making them work. But, if artists adapt their practices to mitigate and hold these companies accountable, the stigma around AI art might change for the better.
Anyway, that is all from me! Have a great day, everyone! π