Midjourney—the Artificial Intelligence Program that Makes Art
Photo is of Jason Allen’s piece of artwork, via DelSignore (2022)
This post was submitted to and published in the spring 2023 edition of Idiosyncrazy, a new undergrad literary magazine at UNC. Idio’s website is https://tarheels.live/idiosyncrazy/about/
It seems like the plot in Zima blue from love death & robots is suddenly coming true; that artificial intelligence is outplaying our best human artists. In August 2022, Jason M. Allen from Pueblo won first place among the emerging digital artists for the 2022 Colorado State Fair’s annual art competition. His work Théâtre D’opéra Spatial (shown above, translated to Space Opera Theater) was not painted, but rather generated using an artificial intelligence program called Midjourney. During an interview, Jason M. Allen admitted to the fact that he has downloaded thousands of pictures with different keywords from Midjourney and then has made selection and modification to the work using photoshop. After disclosing details about the creation of Théâtre D’opéra Spatial, great controversies have been raised in the artist community(DelSignore, paras. 2, 2022).
Midjourney is an open-source artificial intelligence program similar to OpenAI's DALL-E that creates images from textual descriptions. It has the ability to generate pictures or paintings integrating various art styles. It also manipulates existing images by rearranging the objects in the image as well as by adding textures and shadows to them. There are other AI programs that serve similar functions, yet DALL-E by far is the most powerful one—— the majority of people cannot differentiate its work from that of a true human artist. Even so, Midjourney cannot produce images out of thin air; its machine learning uses a model called “diffusion probabilistic models,” a deep learning autoencoder trained to denoise images (noise usually refers to unintended, random variation of brightness/color due to electronic noise) blurred with Gaussian noise, or normally distributed random variations in brightness or color. After learning to reverse the diffusion process, the model would be able to generate new natural images from a random noise image.
So, is the AI Midjourney a machine that recognizes the patterns in existing art pieces and copies human art style or is it considered an independent artist with originality thus entitled to recognition?
Some argue that the AI cannot generate art independently and instead characterize it as only a piece of tool that envisions human artists’ boundless imagination. Users must input a few keywords/descriptions for the picture they wish to create before downloading the product. Obviously, using an AI to create pictures with only verbal instructions is revolutionary compared to all previous forms of visual art media such as Photoshop and Procreate. The most controversial part of this new piece of tool is that it divides the process of creating art into execution and imagination, where one part is completed by an algorithm and the other by a human user. Moreover, users of AI Midjourney improve their aesthetics instead of practicing artistic techniques such as spatial arrangement, color selection, the use of brushstroke, etc. In other words, an artist's advancement and progress made using Midjourney is not translatable to any other forms of art media and perhaps vice versa. In fact, Jason M. Allen of Pueblo, West Colorado was never trained as an artist; he is a game designer.
It's true that most ideas are built on each other, and the use of reference is common for all artists. The creation of art requires inspiration and some levels of trial and error. Overtime, the definition of art has evolved to be more comprehensive and inclusive. For example, the emergence of contemporary and performance art expanded its purpose beyond mere aesthetic pleasure. These art movements raise profound social and humanity paradox for their audience to interpret and ponder. However, AI such as Midjourney makes the detection of plagiarism significantly more difficult. Instead of directly tracing the work of one artist, a new piece of art can be produced effortlessly by mixing multiple ideas with the help of AI.
Without the use of proper data source, it’s very likely to have overfitting, in which a machine learning model has become attuned to the dataset on which it was trained so closely that it loses its applicability to any other dataset. When used without sufficient data source particularly, the images created by the AI would not pass the plagiarism check. On the other hand, immature AI such as Novel AI uses data sources such as Danbooru, itself contains many unauthorized pictures, undermining the resulting AI’s integrity.
The development of technology cannot be halted, nor shall it be disrupted. When there are ethical issues revolving around a technical advancement, the restrictions should be specific to its application. In other words, the ethics of AI generated art is in regard to immoral users. Artificial intelligence programs such as DALL-E provide a window for inexperienced artists to express themselves without spending unnecessary effort on learning the seven basics in art (color, composition, value, form, brushwork, and perspective). Though some artists see the development of personal style and innovation through these seemingly tedious exploration and navigation of basic techniques. On the other hand, plagiarism detection programs specific to DALL-E and legal protection need to be implemented before this new technology continues to serve those in need.
References
DelSignore, P. (2022, September 5). Ai art wins competition and sparks controversy. Medium. Retrieved December 6, 2022, from https://medium.com/mlearning-ai/ai-art-wins-fine-arts-competition-and-sparks-controversy-882f9b4df98c
Ellison, S. Esq.( 2022, August 29). Who Owns DALL-E Images? Retrieved December 6, 2022, from https://www.findlaw.com/legalblogs/legally-weird/who-owns-dall-e-images/#:~:text=Even if you could own,Use call "generations"
The New York Times(2022, September 02). An A.I.-Generated Picture Won an Art Prize. Artists Aren’t Happy. Retrieved December 6, 2022, from https://www.nytimes.com/2022/09/02/technology/ai-artificial-intelligence-artists.html?smid=url-share