AI Art - facts and my views

AI Art - facts and my views
Photo by DeepMind / Unsplash

Recently AI Art has been the centre of controversies on the internet, labelled as the "2023 equivalent of NFTs". In this post I'm going to debunk common myths, misinformation and give you my views on this buzzing topic. In this article I will be using beginner words so that everyone reading can have a good understanding.

Before reading this, it's worth noting my views on AI art may strongly differ from yours, as I am more strongly on the side of research and innovation. Below is a screenshot of how I endorse the usage of AI art.

Is this any different from the past decade?

In the past decade, it's not a secret that the way we've made art has fundamentally changed. In the 1980s, Photoshop was born, which was widely adopted and changed the art community forever.

This has happened in every industry; Sewing machines, music DAWs, and any new technology has shaped the way not only the creative industries, but everything has been created for a while. Some view AI-art as a tool for artists, while other see it as detrimental to the future careers of artists.

Scarcity of real art

If lots of art in the future is AI generated, there would be a huge market for human-made art. As someone that has art in my house, I do not intend on ever removing it to replace it with something created by an AI.

If AI-generated art becomes more prevalent in the future, it is likely that there will still be a significant market for human-made art. This is because many people value the unique perspective, creativity, and emotion that human artists bring to their work. Human-made art can capture the essence of humanity in a way that AI-generated art may struggle to replicate.

In addition, many art buyers may value the physical presence and the history of human-made art, as well as the human connections that can come with it. Human-made art often has a story behind it, which can add an additional layer of meaning and value to the piece.

Furthermore, the process of creating human-made art is often seen as an expression of the artist's individuality and creativity. The act of creating art can be a deeply personal and fulfilling experience, both for the artist and the viewer. Human-made art can reflect the unique perspective of the artist and their experiences, making it a valuable and meaningful part of our cultural heritage.

Adjust to the new tech

When digital audio workstations (DAWs) were first invented, they were initially met with some scepticism and resistance from some musical artists. Some felt that the use of technology would lead to a loss of the human touch in music, while others feared that it would make it too easy for anyone to create music without any real musical talent.

However, as DAW technology evolved and became more sophisticated, many musicians began to embrace it as a powerful tool for creating and recording music. Today, DAWs are widely used by musicians and producers at all levels of the music industry, from home studio enthusiasts to major recording studios.

In fact, many musicians and producers now see DAWs as an essential part of their creative process, enabling them to quickly and easily experiment with different sounds, arrangements, and production techniques. DAWs also offer musicians and producers the ability to collaborate remotely, making it easier to work with other musicians and producers around the world.

I feel that this is the same situation as AI art, and artists will slowly have to keep up or stay to traditional methods and work physically, rather than digitally.

Art theft - The resolution?

All AI models must be trained on a vast dataset of images to be functional, some taking up to tens of gigabytes. This can be seen as problematic for many people, especially artists, as their art and life's work could be part of a massive dataset.

I was thinking about how to combat this, and came up with the solution of only using public domain images. Surely someone has done this before!

The public domain dataset I tried (clean-diffusion-2-0-poc.ckpt)

Yeah.. Not so great. Within a few months or years, I hope to see vast increases in quality of these datasets. I used the prompt "A surfboard on a beach, anime style" to generate this image. Below is the output of SD-v1.5.

Ah... so much better.

But how did we teach the AI in the first place?

Training models

To train an AI art model, a massive dataset of images is needed to provide the system with enough examples to learn from. This dataset is fed into the AI system, which analyses the images and uses them to create a mathematical model of the patterns and features that define each image.

Once the model is created, the system can use it to generate new images that are similar in style or content to the original dataset. The more images the system has to learn from, the more accurate its output will be.


Q: Can AI art models create truly original works of art?

A: This is a subject of much debate in the art world. While AI art models can certainly create interesting and visually appealing works, some argue that true artistic creativity requires human consciousness and intentionality that AI systems simply do not possess.

Q: How can an AI system create art if it doesn't have emotions or creativity?

A: While AI systems don't have emotions in the same way humans do, they can be programmed to recognize patterns and generate new images based on those patterns. Whether this counts as "creativity" is a matter of debate, but it can certainly produce interesting and aesthetically pleasing results.

Q: Can an AI art model really replace the "human element" in art?

A: While AI art models can create interesting and unique images, they don't have the same emotional depth or personal touch that a human artist can bring to their work. It's unlikely that AI will ever fully replace the human element in art.

Q: How can we trust an AI system to make artistic decisions if it doesn't understand art history or cultural context?

A: AI systems can be trained on large datasets that include information about art history and cultural context. While this isn't the same as true understanding, it can help the system make more informed artistic decisions.

Q: Will AI art models make it harder for human artists to make a living?

A: This is a concern that many artists have voiced. While it's true that AI art models can create art that is visually similar to human-created art, there will always be a market for unique, hand-crafted works of art.