Slop work
The Remainder
article
SLOP WORK. In September last year, more than 100,000 people – led by far-right activist Tommy Robinson – gathered in London as part of a nationalist demonstration professing to Unite the Kingdom. The purpose of the protest was to raise awareness around a perceived threat of ‘mass migration’ into the country, with speakers – ironically? – travelling in from all over Europe and the Anglophone world to call for the ‘dissolution of parliament’, as well as policies of ‘remigration’ (mass deportations). Emphasising the supposedly existential nature of their appeal, Elon Musk – appearing via videolink from the U.S. – told the crowd: “You either fight back or you die”.
To anyone travelling throughout the country during the preceding summer this shift in tone and organisational capacities within the far-right could not have come as much of a surprise. Operation Raise the Colours, devised by an anti-migration group of the same name, had called on thousands of ‘patriots’ and ‘christians’ to raise (or spray) the St. George’s Cross on lampposts, bridges and roundabouts throughout the month of August, leading – in some cases – to a total visual transformation of high-streets, highways and other public spaces.
However, this transformation of the physical landscape was not sufficient for what we might call the movement's ambitious ‘evidentiary demands’. The flags might have (at least phantasmally) ‘reclaimed’ territory, but could not depict the threat that supposedly justified their raising: a photograph of a flag-lined high street shows a flag-lined high street, not the ‘invasion’ the flags are meant to repel. This is the gap into which AI-generated imagery has moved. In her LRB essay 1 on the rally and its ‘affective infrastructure’, Claire Wilmot describes a Telegram ecosystem in which deepfakes circulate not as deception but as intentional modes of representation, channels in which ‘users marvel at how well a fake video can represent what they believe’:
“A woman from Norfolk explained that she shared AI images of young white women cowering under the gaze of leering migrants because ‘you can’t photograph them, or they’ll call you racist.’ [...] A Londoner spreading deepfakes of white women saying they don’t feel safe ‘because of migrants’ told me impatiently that everyone knows the videos aren’t real, but I was missing the point: ‘It’s about us showing everyone what’s really happening.’”
*
So, what is ‘really happening’? In the past few years, these images and short videos have come to be known as ‘slop’ or ‘AI slop’ – defined by Wikipedia as:
“...digital content made with generative artificial intelligence that is perceived as lacking in effort, quality, or meaning, and produced in high volume as clickbait to gain advantage in the attention economy, or earn money.” 2
Much like C20th ‘kitsch’ – understood by Clement Greenberg as a ‘low-brow’ cultural reaction to a ‘high-brow’ avant-garde – ‘slop’ is generally viewed by liberal commentators with a mixture of disdain and disavowal. For Greenberg, kitsch was a retreat from modernist abstraction into ‘formulaic’ or ‘mass produced’ modes of cultural representation, aimed at producing ‘unreflective’ emotional responses through illusions of aesthetic or historical depth. He writes: “If the avant-garde imitates the processes of art, kitsch, we now see, imitates its effects…” Think: Soviet Realism (à la Ilya Repin), porcelain plates of Diana, Jeff Koons’ Michael Jackson and Bubbles, or the hallucinogenic jigsaw puzzles of Thomas Kinkade…
These descriptions share much in common with those of ‘slop’ today, and the idea that something ‘fake’ (or ‘deepfake’) might feel more real than reality itself. And much like ‘kitsch’ (which emerged from a German expression for ‘trash’), ‘slop’ is a word of hazy origin – initially attested in the 16th–17th century, first as a term for a ‘muddy hole’ (oink, oink), and later used as a verb ‘to spill’ or ‘splash’ something carelessly.
Yes, everything about ‘slop’ seems like an accident – it appears no one ‘really means’ what it is they produce using generative AI, or ‘really believes’ the results. As Rob Horning writes, in a recent article:
Generated video caters to an audience’s desire to see their beliefs in the form of evidence, and consume what they already believe as if it is being irrefutably established by real events. They visualize simple explanations for us so that we don’t have to put effort into them or take responsibility for them. These kinds of clips allow us to experience such beliefs as content without having to “really” believe in them. 3
If Nazi propagandists turned to kitsch as a kind of ‘realism’ – able to reflect an uncomplicated image of Aryan life back to itself (as Greenberg writes) – contemporary fascists turn to ‘slop’ as a kind of ‘evidence’, imitating (through the use of Sora, Seedance and other tools) the aesthetics of vox pops, CCTV and LiveLeak footage.
Of course there is a sense that – for the sensible and aesthetically ‘literate’ readers of this column – this is simply an issue of people ‘making a mistake’, of literally ‘miss-taking’ a fake image for a real one. However, as Claire Wilmot writes in her LRB piece, “...part of the misunderstanding of the deepfake threat stems from the idea that it is a problem of bad information, rather than a problem of desire…” Perhaps it is the case (spoiler: as it always has been) that people in fact want to be deceived? That self-deceptive imagery – as in kitsch – can provide what Wilmot calls an ‘illustrative diagnosis’ of what is commonly held to be, or represent, the truth (but cannot, for whatever reason, be directly observed).
‘Slop’ – in a way that is perhaps distinct from ‘kitsch’ – is an interesting word for the reason that it implies (both semantically and etymologically) something accidental or careless. It absolves those who engage in its production, dissemination and interpretation (whether that be critical or otherwise) of responsibility. And it does this by design – as Horning writes: “Slop’s dismissibility is part of its appeal.” Slop’s danger, as with kitsch, and as Horning continues, “...[is] not in persuading us of something nonfactual but in eroding [a] sense of collective responsibility for shared information.”
*
The transition of ‘slop’ from noun to verb and back again is telling; slop – like mud – appears all around us, seemingly as a natural feature of the landscape. Ultimately this is a feature of its linguistic make-up. And yet we know, of course, that ‘AI slop’ is in fact a product of very conscious decisions made by people in both Silicon Valley and, most likely, that once great ‘Garden of England’ (Kent). And so here starts my petition for the revival of a term which might make such intentionality apparent: ‘slop-work’.
Slop-work was a term widespread in the textiles industry of Victorian England, emerging in the mid C19th to define mass-manufactured and poorly produced articles of clothing – slop-garments, slop-suits or simply ‘slops’ – as well as the process of their making. As Wikipedia writes under the article ‘Slop (Garments)’: “Slop was made by slop-workers and sold in slop-shops by slop-sellers”; this litany of ‘slop-’ terms highlighting the disdain with which those who engaged in these practices – most often early capitalists exploiting largely female pieceworkers – were met by the general public.
However, as industrialisation moved to the colonies of the Anglophone world, and as the textile business at home ‘advanced’ into ever-more automated forms of production, the term ‘slop-work’ fell out of use – presumably, in part, because the making of slop was no longer so visible (its products, of course, were still widely consumed). Ultimately, this was (is?) an abdication of responsibility – in the Western world – for a fundamentally unjust form of material production, which required (requires?) the disavowal of its commercial and cultural output: i.e. cheap clothes. Slop-work begets slop-work – but importantly, this is conceptualised as work, as labour, and as such can be entered into a well-worn(!) political vocabulary and cultural discourse.
Slop does not spill itself. Like the Shein dresses and boohooMAN t-shirts that litter the U.K.’s charity shops, the ubiquity of ‘AI Slop’ disguises the process of its manufacture. Slop is ‘generated’ – not by virtue of a magical technical process (generative AI), but rather through the deliberate labour of groups and individuals with particular outcomes in mind.
This article was written for a monthly column in the Sticky Fingers monthly mailout