Blog: Can artificial intelligence create art?
In October 2018, the (supposed) first work of art created by artificial intelligence, was sold at auction for $430,000 (£335,000). An obscene amount of money by most people’s standards, and the controversies around the coding of the AI are something to be debated at a later date. But as a Graphic Designer, I can’t help but feel that this is both a monumental step forward, and a grim look into what might become of my job in the years to come…
This is not the only instance of AI producing creative content, there are now AI logo creators that can produce a business logo in only 30 minutes based on your likes, dislikes and company ethos. And the reviews are mostly positive of this. I too am mostly in favour of this new mode of design — it is quick and easy, and especially for companies just starting out it can be a great help! However, if you wanted a logo with more meaning behind it; would AI be able to deliver this without an understanding or conversion of human semiotics behind it? And by extension, if AI creates art based on an algorithm, does this count as art if it is not produced for any form of visual communication. Does the goal of an art piece matter? Or is it the consumers responsibility to imply intersemiosis (O’Halloran et al., 2012)?
It is important to note that whilst we use the terminology: artificial intelligence; it is not yet fully autonomous, and as such, is simply an electronic production of a human’s vision. The public is quite confused by the marketing of ‘AI’ because they are fundamentally confused about what AI is. A true form of AI would be that which starts as a string of code, but evolves and grows on its own, learns and writes itself based on its own experience, until the original code is rendered redundant. Once an AI can make its own autonomous decisions — then these questions would really come into play.
Visual communication and multimodal integration:
Increasingly more time is being devoted to visual aspects of communication by researchers (Barnhurst, 1994), and this includes multimodal messages and how they combine various modes of representation (Griffin, 1992; Kress & Van Leeuwen, 1996). This attention also covers the relationship of visual design and textual design (Bucher, 2007).
When discussing visual communication and multimodality it’s important to know how much of an impact it has not just on art, but our day to day lives. It is dominant within our subconscious and yet there is still a lack of studies surrounding how people interact with visuals and media messages — which makes it slightly harder to reach a conclusion as to whether AI can create art as we do not have enough empirical evidence on how we, as humans, interpret art. It begs the age-old question; what is art? And does the receivers’ perception matter more than the original production?
Multimodality describes approaches that understand communication and presentation to be more than about language and semiotics; and which encompass the full range of communicational forms that people use e.g. Images, Gestures, Posture etc. And the relationships between them. Today messages in the media, and throughout art and imagery, combine a high degree of semiotic complexity involving various modes in different ways.
Multimodality within art:
The multimodality within art can be analysed from both a production and reception perspective (Bucher, 2012; Holsanova, 1999). Production refers to the interplay between various modes, their contribution to the message content and their output in order to create a certain desired effect. This is often referred to as Intersemiosis. Media producers often form their message (art, music, text etc.) on how an imaginary or hypothetical recipient would perceive the material. If there is no message behind the art, does this detract from its value, and therefore would art produced by AI be notably different from that of a human’s art, simply due to the emotive process behind it?
The composition of multimodal documents and its potential for meaning-making have been discussed in the social semiotic tradition (Kress & Van Leeuwen, 1996; Holsanova, 2006), as well as rhetorical tradition (Bateman, 2008). And the reception of visual communication is closely connected to the consumers ability to select, attend to and process information. It is the recipients that ultimately choose among the available information: What they want to explore, mentally process and interpret, in what order, and how deeply. Art is in the eye of the beholder. Therefore, it would not matter if a form of media were created by AI or a human, it would only matter if the visual communication made sense to a consumer, and they interpreted said media as art.
We can deduce then, that recipients play an active role in the interaction with multimodal messages. The perception and interpretation of visuals can be seen as an interactive meeting between the recipient, the multimodal message and the situational context. This means that different people will perceive the same visual differently, and attribute different meanings to it. What one consumer may call art, will not necessarily be what a different consumer would consider to be art. AI created art will be of value to some people, and not to others.
Art, or other signifiers, are formed by the producer. However, the meaning is formed by the receiver?
The way that consumers receive, understand and remember the context of the multimodal messages is guided by visually salient and semantically relevant aspects. Ergo, a recipient’s perception is altered by not only the colour and format etc. of a visual, but also the recipient’s own life experiences (Boeris & Holsanova, 2012).
Previous research has shown that visual attention allocation is often influenced by pre-existing knowledge and expertise (Van Gog et al., 2005). For instance, experts in a subject field will use wildly different analysis and exploration methods to those of complete novices. This too impacts the meaning-making, symbolic, indexical and iconic design of multimodal messages. So would AI be able to adjust its logo design, or artwork to hold gravitas, based upon the pre-existing knowledge of their target audience?
The task and viewer’s goal also have an impact on their visual exploration behaviours (Yarbus, 1967). If a consumer goes to an art gallery, they are in a mindset to see art. So even if in other setting a visual may not be considered art — in the correct setting and from the correct consumer, anything could be! And the amount of prior knowledge of the consumer based on life experience may determine how much insight and interpretation they can receive from any given input. Given this, there is still a need to study the affective responses to visual communication and multimodal messages. As without looking further into our own understanding of how we process and perceive art, how are we to know what art is, and therefore determine if something created by artificial intelligence, is or is not art?
Art is subjective…
In the end we must conclude that all art is subjective; what I consider to be beautiful and hold meaning, will not be the same for you. It is based on personal perception, my individual differences and expertise. My goal when interpreting a visual will be different than yours, and therefore the context for perception and interpretation of the visuals. The emotional impact of the visuals may be different, as will be the process of meaning making and intersemiosis. The relation between visual attention, meaning attribution and emotion will be different for each and every person. And so, what I consider to be art may not apply to everyone.
If art is defined by the consumer, and not purely the motivation behind the visual communication, what percentage balance would require something to be considered art? And would that allow for AI to create art? As someone may feel a connection to it without there being a personal message behind the work.
In my opinion, the current version of artificial intelligence does not create art, it creates a version of the coder’s interpretation of what art is. Until artificial intelligence is fully autonomous, and learns, and experiences, and decides itself to create art, for me it holds no meaning. Though let’s be honest, if you have $430,000 to spend on a piece of art/not-art, does the meaning really matter that much if you think it’s pretty?