A Question of Artistic Labour
Or why art has long been about ideas, not just physical labour.
<h3 dir="ltr">I often hear people dismiss AI-generated art due to the misperception that "prompting" lacks artistic labour. After all, if you just type out what you want to see and the machine churns an image out based on your words, <em>did you really make art</em>? The mental image of the toiling, struggling artist is so pervasive, that people almost resent the absence of physical labour in contemporary artistic practices. It may surprise most to discover that prompt-based art has actually been around for nearly a century.</h3>
<p dir="ltr">One of my personal favourites is:</p>
<blockquote>
<p dir="ltr">A 6” (15 cm) grid covering the walls. Lines from corners, sides, and center of the walls to random points on the grid.</p>
<p dir="ltr">1st wall: Red lines from the midpoints of four sides;</p>
<p dir="ltr">2nd wall: Blue lines from the four corners;</p>
<p dir="ltr">3rd wall: Yellow lines from the center;</p>
<p dir="ltr">4th wall: Red lines from the midpoints of four sides, blue lines from four corners;</p>
<p dir="ltr">5th wall: Red lines from the midpoints of four sides, yellow lines from the center;</p>
<p dir="ltr">6th wall: Blue lines from four corners, yellow lines from the center;</p>
<p dir="ltr">7th wall: Red lines from the midpoints of four sides, blue lines from four corners, yellow lines from the center.</p>
<p dir="ltr">Each wall has an equal number of lines. (The number of lines and their length are determined by the draftsman.)</p>
</blockquote>
<p dir="ltr"><br>These are instructions written by Sol Lewitt in 1975, for <em>Wall Drawing 273</em>, an artwork intended to be realised by “a draftsman,” per Lewitt’s own instructions, who is likely a museum worker, or even an intern.</p>
<p dir="ltr"><strong><img src="https://cdn72109500.ahacdn.me/news/upload-1586705043.jpg" alt="Sol Lewitt, Wall Drawing 273, 1975, Image: San Francisco Museum of Modern Art" width="1324" height="1146"><br></strong><span class="custom-span">Sol Lewitt, <em>Wall Drawing 273</em>, 1975, Image: San Francisco Museum of Modern Art</span></p>
<p dir="ltr">As the artist of the artwork, Lewitt’s most direct input is that of creating the instructions. In today’s digital art lexicon, we would refer to said set of instructions as a prompt. In each realisation of the work, the now 49-year-old prompt should result in a geometric, large-scale work of art, likely iterated over time with different results, drawn by different people, with different materials to make the lines, as the work is revisited, or sold to another institution or collection.</p>
<p dir="ltr">Lewitt and plenty of other artists have historically delegated their labour, and there was a time when the contemporary art world lauded them for it. One could even argue that Duchamp’s found sculpture, <em>The Fountain</em> (1917), was a result of delegated labour, and an unintentional one at that, since it was utilitarian, rather than artistic.</p>
<p dir="ltr">So how is writing a prompt for Midjourney or DALL-E to dream something up all that different than what Lewitt did 50 years ago? Or, if it isn’t, why are we hellbent on questioning the artistry of the prompter and the artistic legitimacy of the outcome?</p>
<p dir="ltr">Instructions were a critical part of many artistic experiments in the 20th century, most notably during the Fluxus movement, which emphasised process over end result. There was a performative focus that insisted on framing the transient act of making art as art itself. Although not a Fluxus artist himself, Lewitt is irrefutably part of the Western art canon.</p>
<p dir="ltr"><img src="https://cdn72109500.ahacdn.me/news/upload-1145181531.jpg" alt="Robert Rauschenberg, White Painting [three panel], 1951, Image: San Francisco Museum of Modern Art" width="1661" height="1124"><br><span class="custom-span">Robert Rauschenberg, <em>White Painting [three panel]</em>, 1950, Image: San Francisco Museum of Modern Art</span></p>
<p><strong> </strong>Pre-dating Lewitt by two decades, Robert Rauschenberg has white paintings produced in the 1950s that he had in fact delegated to be produced by others per his instructions, with both the act of delegating and his absence being part of the artistic process. Famous Fluxus artist John Cage’s <em>0’00” </em>(1962) has no musical notes and is instead comprised of a single phrase of instruction: </p>
<blockquote>
<p dir="ltr">In a situation provided with maximum amplification, perform a disciplined action.</p>
</blockquote>
<p><strong> </strong>There are extraordinary examples of how different artists, musicians, and orchestras interpret this prompt. Not only is each interpretation valid, but more importantly, the more people attempt to realise this simple phrase, the bigger and more interesting the artwork becomes. If anything, it is kept alive because of how intriguing and exciting it can be to reinterpret the instruction.<strong> </strong></p>
<p dir="ltr">None of these examples are intended to frame AI-generated artworks as requiring precedent to be artistic. They can be artistic regardless. But hot on the heels of a <a href="https://www.artnews.com/art-news/market/christies-ai-art-sale-augmented-intelligence-controversy-surpasses-expectations-1234734870/">somewhat controversial auction of AI-generated works at Christie’s</a>, context seems necessary. The current Luddite position against prompt-based artworks undermines not only today’s digital creations but the better half of the last century of art. </p>
<div class="interactive-block">691</div>
<p>The misconception that AI-generated imagery is lacking in artistic labour due to its instruction-based process, could perhaps be attributed to a general lack of Art History education. Delegating labour may create the impression that there is no real artwork being made, a scepticism that is in part fueled by the bitter aftertaste of the NFT bubble, where overnight, everyone was a digital artist peddling iterations of pixel-art sketches of bulldogs and the like. The damage the hyper-commodification of digital visuals has done to digital art will take some time to fix in the public psyche. In the meantime, however, it is worth recognising that intellectual labour, though harder to measure, can and does exist in prompt-based AI works of art. The more difficult notion we are wrestling with today, is the painfully ableist question of whether physical mastery of manual tools and/or our bodies is a prerequisite for art to be considered valid.</p>
<h3 dir="ltr">In other words, does the ability to eloquently and strategically verbalise one’s imagination make one an artist?</h3>
<p dir="ltr">There are countless ways to generate visuals based on prompts in DALL-E and Midjourney, two of the most popular visual AI engines that are publicly accessible today. Learning them takes time, as does honing the image and communicating with an intelligence that does not quite yet understand reality. Sometimes, the artistry lies in knowing which of the generated images to select, the way a film director might choose a take of a scene, or a photographer might choose a frame out of 50 shots of the same subject.</p>
<p dir="ltr">Sure, these algorithms mine the Internet for visual references, which include imagery not necessarily done by the artist writing the prompt, but unless we banned collage-making while no one was looking, AI is not doing something that every teenager with access to print issues of Vogue and Time and Blue Jean hasn’t done in the 80s or 90s (sorry Gen-Z, your subscriptions are digital).</p>
<p dir="ltr">Language, more than anything else, holds the key to understanding the artistry and labour that goes into AI-generated artworks. When generating an image of the sky, typing in “blue” is not the same as “cerulean,” and understanding the difference goes a long way. But the more exciting part of it all is that even if you type in “cerulean,” no two images can be the same.</p>
<p dir="ltr">Cage and Lewitt would have had a field day.</p>
<p> </p>
