ARTificial: AI art, its applications and implications

Trinity News explores the phenomenon of AI art, and its place in the art world

AI Art is the latest trend taking hold of social media and the art world. From creating novel pieces to novelty TikTok filters, the development in this area of machine learning may change our understanding of creativity. However, before we hand our culture over to disembodied machines, we must ask what is AI and what are the consequences of its implementation for human society at large?

AI (artificial intelligence) is a particular area of computer science that centres around building machines that can mimic and mirror human behaviours. This is done by feeding data into particular algorithms that enable them to produce human-like responses to stimuli. In this sense, AI machines can learn human behaviours and subsequently produce art by being fed a particular algorithm and the right visual aids. Through this, AI machines are able to produce art through their ability to simulate a human brain and mimic its behaviours. 

“Engineering, mathematics, and creativity combined to expand the boundaries of computer science, with artificial intelligence representing one of the field’s most groundbreaking innovations.”

According to Artland Magazine, the first attempt to figure anything close to artificial intelligence was attempted by Ada Lovelace in the 1840s with the Analytical Engine, which is now considered to be the first computer. From this, engineering, mathematics, and creativity combined to expand the boundaries of computer science, with artificial intelligence representing one of the field’s most groundbreaking innovations.

The major technological achievement that enabled AI to be able to generate ‘original’ works of art came about in 2015, when Alexander Mordvintsev created the DeepDream database. Originally, the system was created to understand how Artificial Neural Networks in AI worked. In a 2015 blog post, Mordvintsev explained that “we train an artificial neural network by showing it millions of training examples and gradually adjusting the network parameters until it gives the classifications we want,” but added that “one of the challenges of neural networks is understanding what exactly goes on at each layer.” In an attempt to dissect the inner workings of these networks, DeepDream was created, allowing the programmer to feed the AI certain images and parameters, noting the image that the program produced in response. By adjusting the parameters and stimuli, then examining how the image changed, Mordvintsev and his team tried to understand how the AI worked and consumed data. The result was a variety of original, psychedelic images, which led Mordvintsev to “wonder whether neural networks could become a tool for artists a new way to remix visual concepts or perhaps even shed a little light on the roots of the creative process in general.” By feeding AI data and getting it to produce images, the AI was able to create new artworks. In a way, this mechanism displays AI’s ability to understand the natural human tendency to create through its programming and mimic this in its production of art.

Due to its ability to produce new and interesting images, DeepDream indicated the ability of AI to be used in artistic pursuits, acting as a springboard for the development of AI specifically intended for producing art. Today, this idea has become thoroughly commonplace, since a quick Google search displays thousands of results for AI art generators. The interest in AI art is also high across social media platforms, with #AIart having over 2 billion views on TikTok, while the platform offers its own AI filters for users to try themselves. One recent TikTok trend saw people turning themselves into Manga characters using an AI filter, testing what the AI would pick up or transform. This particular filter currently has 113 million videos of people trying it out posted to the platform.

Beyond providing entertainment as a glorified 2014 Snapchat filter, AI generators have also found other uses in the art world, as AI programs have been employed to authenticate pieces of art. The company Art Recognition has spearheaded this idea, offering the ability to authenticate art from a single photograph of the piece. This is achieved by collecting a dataset of works known to be made by an artist, and feeding this data to an AI so that it can use machine learning to identify the artist through their “brushstroke, object placement, use of colour, and high-level compositional elements.” This follows the same principle as AI art creation: feed the AI stimuli and parameters so that it can break down these elements and analyse them. 

“If an artist uses an AI-generated image, is the work theirs?”

It has been argued that AI can serve to help artists within the creative process. An AI could generate an image from what the artist feeds it, allowing the artist to then interpret or replicate that image in more traditional mediums. However, here arises the issue of ownership and intellectual property surrounding the entire production of AI art: if an artist uses an AI-generated image, is the work theirs? When artists create their own original works using more traditional or tangible mediums, the work is automatically understood to be theirs under copyright law. However, because AI art is such a recently-developed concept, its regulation is not explicitly encompassed under any existing copyright law documents. The United States’ Copyright Act notes that, in order to obtain copyright protection, the art must be “an original work of authorship”, which is usually interpreted to exclude AI-generated works. In this sense, AI art can have no legal owner, suggesting that no work of AI art is truly original since the algorithm’s pieces wholly rely on synthesising other existing artworks. The Center for Art Law explains that “there may be infringement claims on the final [AI] image based on copyrighted artworks inputted into the AI at the time of machine learning which may infringe the rights of copyright holders.” Thus, the mechanisms of AI art programs can actively inhibit its users from claiming ownership over any pieces of art that they generate. 

Aside from the legal ambiguities posed by AI art, the phenomenon of computer-generated pieces has highlighted many ethical questions surrounding artificial art. Writing for Forbes Ben Meisner poses two tangential questions; whether AI art can be deemed truly original and whether it can be considered “true art”. Both of these debates must be addressed if AI art is to become more prevalent in artistic and creative spaces. 

In terms of whether AI art is truly original, it cannot be overlooked that AI must be fed pre-existing works in order to teach it what “art” is, let alone produce anything new. In this sense, anything produced by AI will be a derivation or interpretation of what AI understands as art, based on the images it has processed. Thus, it can be argued that AI is unable to produce truly original works. However, can this idea not be reflected back onto human methods for producing art? Whether it is art in the traditional sense or indeed any form of creativity inspiration to create often comes from observing or consuming the works of others. Nonetheless, perhaps inspiration and subsequent creation are forms of human instinct that cannot be replicated by machine learning, underpinning what art is at its most surface-level sense.

“Can we truly attribute the label of art to a piece generated by a machine that mimics human behaviours?”

Meisner notes that “we commonly think of art as those forms of expression that come from someone’s emotions and that we relate to on a human level.” This is another instinct that drives human creativity: emotion. All of art can be viewed as a reaction to an external stimulus and while AI’s main function may be to react to its parameters and data in order to produce an output, it would be a stretch to call this an emotional reaction. Can we truly attribute the label of art to a piece generated by a machine that mimics human behaviours? An AI can potentially mimic emotion, but this is only imitation, so perhaps AI art should be classified instead as an imitation and held separate from art manifested from human inspiration, emotion, and instinct.

It seems that artificial intelligence programs can aptly understand and mimic human behaviours, to the extent that they can produce original works, but surely there are limitations to how much a computer can reflect the human and these limitations must also be placed on the art created. If art is expression, then AI art is truly just a mimicry of expression.

Lara Mellett

Second Year English Studies student at Trinity