<p>Acclaimed digital culture theorist Lev Manovich on what will make AI artistically meaningful</p>

Lev Manovich

Towards ‘General Artistic Intelligence’?

<p>Geoffrey Hinton, the ‘Godfather of AI,’ recently predicted that so-called general artificial intelligence could emerge within 20 years. This term refers to a hypothetical future AI that can perform any cognitive task that a human can. This statement begs the&nbsp;question: how long will it take before we see the emergence of general <em>artistic</em>
intelligence? And what should such hypothetical general artistic AI be capable of?</p>
<p>The current more limited paradigm for creating artistic works with AI is known as ‘generative media.’ (Not to be confused with ‘generative art.’) Trained on hundreds of billions of images and their text descriptions, this more limited AI learns to predict a probable image when you give it some text. The images it generates did not exist before, and depending on the prompt you provide, they can look either stereotypical or be quite unique. And if the artist is using Midjourney (currently the most advanced AI image tool), generated images are often aesthetically pleasing. They can have perfectly balanced compositions, beautiful color palettes, and lines and shapes forming rhythmical patterns.</p>

<p>So, while we can endlessly debate if AI is ‘creative’ or not, it is already fully ‘professional’; technically more accomplished than many art students and adult artists. Having taught in university art departments for 20 years, I am also certain that an applicant with a portfolio of only AI-generated images and an artist statement written by ChatGPT will be admitted to most art programs. In fact, this already is a real issue for art faculty who serve on admission committees.</p>

<p>Another way to interpret the aesthetics preferred by AI image tools is to use the term ‘kitsch.’ Kitsch emerged in the art markets of Munich in the 1860s and 1870s as a term to describe inexpensive and popular pictures. According to Tomas Kulka’s analysis in his book <em>Kitsch and Art</em>
(1996): ‘Kitsch depicts a beautiful or highly emotionally charged subject; The depicted subject is instantly and effortlessly identifiable; Kitsch does not enrich our associations related to the depicted subject.’ Kitsch, in other words, is melodramatic, shows only stereotypes, and lacks originality. This, to me, is an excellent description of the default images produced by AI tools. It is possible to make them not kitsch looking, but it takes time and prior experience in a field such as illustration or photography.</p>

<p><strong>The art of the copy<br><br></strong>AI image generation can be referred to as ‘the art of copying.’&nbsp; There are four distinct types of copying going on. Firstly, you can endlessly repeat the generation process by running the same text prompt, hoping to see improvements from the original image. Many artists reported using the same prompt thousands of times until they got the desired images. Secondly, there are websites that collect millions of prompts and resulting images, so you can directly copy any such prompts, and then modify them. Thirdly, in the Midjourney default mode, you can view in real time thousands of other users entering their prompts and the images they produce, and others can view your prompts and images as well. Finally, AI image generation is ‘the art of the copy’ in the sense that many prompts refer to popular video games, Hollywood films, animation studios, and popular illustrators and artists. </p>

<p>What differentiates this AI-enabled ‘art of the copy’ from, for example, the millions of amateur artists creating copies of anime characters and uploading them on ArtStation and DeviantArt? One new aspect is the mechanism of copying. Instead of tracing over existing images by hand, as artists used to do, you can now copy text prompts which act as a ‘code’ for an AI image tool. Another new aspect is the modularity of the prompts. You can copy only words describing the aesthetics, composition, a feature or character, type of lighting, and so on, and then remix it with your own content description.<br>
</p>

<p>Before we dismiss vernacular AI visual culture as completely derivative and unauthentic, consider how important copying has always been in human cultures. Historically, image making was often about making copies or variations. For example, Bruegel's family produced approximately 127 copies of his painting <em>Winter Landscape with Skaters and Bird Trap</em> (1565). According to political economist Krzysztof Pelc, ‘half of all commissioned artworks in the 16th century were copies of originals.’<br>
</p>
<p>In my opinion, museums that present the history of art through the lens of Modernism and emphasize uniqueness and originality do not allow us to fully appreciate the importance of copying historically. You can enter a room devoted to art from a specific time and place yet see a multitude of contrasting styles and approaches. This sampling bias toward uniqueness does not reflect the culture of copies and minor variations. The universe of AI imagery created by amateurs today reminds us of how cultures have always operated: through constant imitation and small modifications.</p>
<p>In conclusion, I want to come back to my opening question: what do we want to see in a hypothetical general artistic AI? While image generation tools will continue to improve, I want an AI that does not make any art on its own. Instead, this AI would use algorithms to analyze an artist's works and provide them with insights and new ideas. It could visualize changing patterns in the artworks made over time in subject and form and suggest possible directions for experiments. It could identify which techniques the artist employs frequently and infrequently and encourage the use of more uncommon ones. In other words, even though such AI wouldn't automate the creation of art (and why would we want this anyway?), it could offer artists something more valuable: intelligent feedback and even dialog.</p>

<p>One might wonder if the march of digital innovation over the past several decades has altered Arcangel’s approach to artmaking. ‘On a large scale, it hasn’t,’ he states, ‘because for me, my work is rooted in a feeling. And that hasn’t really changed since I was a kid.’ We agree that tech may change what art looks like, and what tools artists consider primary, but that ultimately art is of a different order to tech. ‘The two have always evolved in tandem,’ we offer, thinking about how changes in infrastructure alter surface effects but not the human spirit. ‘Yeah, it’s gonna be cool,’ Cory replies, signing off, ‘I think it’s going to be a wonderful mess.’</p>
<hr>
<p><a href="https://www.artbasel.com/catalog/artist/8778/Cory-Arcangel" target="_blank">Cory Arcangel</a> is represented by <a href="https://www.artbasel.com/catalog/gallery/1188/Lisson-Gallery" target="_blank">Lisson Gallery</a> (London, New York, Shanghai, Beijing, Los Angeles). He will present an artwork in the Unlimited sector of Art Basel in Basel 2023.<br></p>
<p>Caroline Busta and Lil Internet are cofounders of <a href="https://newmodels.io" target="_blank">New Models</a>, a media platform and community addressing the emergent effects of networked technology on culture. Busta was previously editor-in-chief of <em>Texte zur Kunst</em>, and an associate editor of <em>Artforum</em>. Lil Internet has worked as a video director (Beyoncé, Diplo, Vogue, Adidas), music producer (Azealia Banks), and pop media theorist. They are both based in Berlin.</p>
<p>All photos by&nbsp;Tonje Thilesen for Art Basel.</p>
<p>Published on May 17, 2023.</p>

title_style_2

<p></p>
<p>The rising star transcends identity politics with words, scrap metal, and extracted gold</p>
<p></p>

Stories

Unholy motors: Meet Selma Selman, the Bosnian artist turning machines into art

<p>This panel examines how recent, much-publicized, advances in AI image-generating programs such as Midjourney and DALL·E 2 are shaping both artists’ and the publics’ imagination</p>

Conversations

(Co-)Creating with AI: The Artist’s View

<section></section>
<section> <section><main><div><div><section></section><section> <section><main><div><div><section></section><section> <section><main><div><section></section><section><section><main><div><div><section></section><section><section><main><div><section></section><section><section><main><div><section></section><section><section><main><div><div><section></section><section><div><div><section><main><div><div><div><p>As he prepares to unveil a new site-specific commission for Art Basel’s Parcours sector, the artist reflects on the challenges of creating art for public spaces<br></p>
<p></p>

</div></div></div><section></section></main></section></div></div></section></div> </div><section></section></main></section></section></div><section></section></main></section></section></div><section></section></main></section></section></div> </div><section></section></main></section></section></div><section></section></main></section></section></div> </div><section></section></main></section></section></div> </div><section></section></main></section></section>
<section> <section><main><section></section></main></section></section>
<section></section>

How Hank Willis Thomas became one of America’s most prolific public sculptors

Acclaimed digital culture theorist Lev Manovich on what will make AI artistically meaningful

Welcome,

Discover more related content below: