AI ART

Return to Menu

What is AI?

AI stands for Artificial Intelligence and, as the name suggests, describes a machine that shares the same capacity for intelligence as a human does. Its difficult to define intelligence but as a baseline - the Oxford Dictionary describes intelligence as: the ability to acquire and apply knowledge and skills. When it comes to AI art, the idea is that a computer program can construct an image entirely on its own that will be new, unseen and fully unique.

The History of AI Art

Game Of Life

Game of Life is a cellular automation system created by John Horton Conway in 1970. The program works by using the initial input to progress the image via individual pixel “cells” that follow a set of rules. The rules include:
-Any live cell with fewer than two live neighbours dies.
-Any live cell with two or three live neighbours lives on to the next generation.
-Any live cell with more than three live neighbours dies.
-Any dead cell with exactly three live neighbours is brough back to life.

AARON

In 1973 Harold Cohen created a series of computer programmes he named AARON, which could create original abstract images. Cohen used various machines to transform the digital AI art into physical art pieces. The machines used ink and fabric dye to transfer the designs onto canvases. At some point, Cohen even opted to paint the designs himself by hand. By the 1980s the program was taught to creat more concrete shapes that resembled actual objects and even people, evolving the once abstract images into more detailed, complex ones. 		Image source: researchgate.net
Image source: michaelhoweely.com

DALL-E

In 2021, OpenAI - a company owned by Elon Musk - released DALL-E to the public. The program made AI art creation simpler than ever, all users were asked to do was create an account and input whatever terms they please into the generator search bar; everything else the program did itself. As likely expected, the program quickly became closely integrated with meme culture, particularly on platforms like twitter where users would compete with each other to determine who can make the most bizarre and nonsensical image. I chose DALL-E as the program to experiment and try to make my own AI art with.
Artificial Intelligence Generate

Starting with the obvious, my first search was the phrase ‘Artificial Intelligence’. Not entirely sure what my expectations were from this search, however I must admit the generated results were quite disappointing. The images were very basic and generic, making me almost certain that if I were to enter the exact same prompt into Google, the search results would present me with identical images. This is not a fault of the program as much as it is the lack of detail in the prompt itself; OpenAI stresses the importance of using as much detail as possible when describing the desired image, e.g. making note of all of the elements as well as the overall style in order to achieve optimal results. This search simply proved to me that the ‘scary’ AI does not indeed have a mind of its own, at least, not as much as I initially thought.

Intelligence Generate

I followed my previous search with a similarly simple prompt to get a baseline of the program's understanding. The results were equally as lacklustre.

The Meaning of Life Generate

To challenge the program even more I decided to give it the ultimate question: what is the meaning of life? The result: a series of photorealistic close-up images of still nature. Although extremely simple in their format and exposition, the images convey an almost Romantic perspective of life. Each photo focuses on seemingly menial aspects of nature - elements of the surrounding world we tend to overlook and ignore in favour of deeper and more complex ideas. Perhaps I am looking too deep into these simple images, but I must admit, I was left speechless looking at these photos; I struggled to understand how an AI could have such an intricate understanding of life.

A victorian painting of a cat king dressed in royal robes and a crown Generate

I found that painting styles produced the best quality images and experimented with the limits of the program. My favourite search was for a Victorian painting of a cat king. The quality of the generated images was exceptional - almost that of paintings that belong in a real museum. I was curious to see just how original the images were and Google the exact same prompt. The results, although giving me what I asked for, were nowhere near the quality and accuracy of the AI generated images. There was a sort-of authenticity to the generated paintings, making me believe that if I looked closely enough, I could find the texture of the non-existent paint brush strokes.

Baroque painting of a distressed watermelon Generate

I experimented more with paintings and came up with this result. The quality of the images was equally as exceptional. Although I found the program's interpretation of the word ‘distressed’ interesting, it was not what I initially intended when searching, so I repeated the search with a clarification.

Baroque painting of an emotionally distressed watermelon Generate

The results for an emotionally distressed watermelon produced these monstrosities which, although great in quality, left me sleeping with the lights on that night.

Busy Streen in Paris Generate

Using DALL-Es newest technology called "Outpainting '' I extended an existing image I had in my camera roll. The image I used was of a Parisian landscape, hence my prompt described a "busy street in Paris". For the most part, the image turned out pretty okay, especially if you observe it from a distance; however, looking at the image more closely uncovers many inconsistencies and obvious errors. Additionally, one half of the sky ended up being completely different, which could possibly be an issue of having to expand the image in pieces, one at a time.