Who is watching over the city of lights?

As I walk towards a well-respected arts institution in the heart of Paris, I pause to observe my surroundings. What I discover is a disturbing anomaly. Poorly concealed within the structure of a lamp post – a remnant from the 19th century, where, until recently, one would expect to find a buzzing, yellow-hued sodium vapour bulb – I notice a tight cluster of dome-shaped surveillance cameras.

The history of public lighting in Paris is both long and fascinating. Detailed accounts of the early (and failed) attempts to light the streets stretch back to the reign of Saint Louis in 1258. When functional lighting arrived in Paris, it did so in the form of the Cardan oil lamp in the 17th century. Two hundred years later, Paris witnessed the introduction of gas-powered streetlights, followed by the advent of electric lights at the end of the 19th century. More recently, much has been said about the shifts from those quintessentially Parisian warm-yellow bulbs to whiter LED lighting. However, an even more recent change, largely absent from the historical accounts of lighting systems in Paris, is the slow and creeping infiltration of surveillance cameras. Today, a simple online search will reveal the various shapes and sizes of cameras that can be installed within preexisting lamp posts. Betraying some shade of genuine virtuous ingenuity, these contemporary solutions are not only constructed with an eye towards a future with even scarcer energy resources – solar-powered streetlights with CCTV cameras installed directly inside the light itself – but one in which citizens are increasingly surveilled, controlled and mistrusted.

My interest in the Parisian light fixtures dates only to the last few months, during which my work focused on the development of video surveillance algorithms. Yet, public lighting has long been linked to surveillance. As Jonathan Crary tells us: “the broad deployment of urban streetlights by the 1880s […]” not only “expanded the time frame and thus the profitability of many economic activities” but also served to “[reduce] longstanding anxieties about various dangers associated with nocturnal darkness” (Crary, 2013). Where there was light, there was both an increasingly lengthy workday and a perception of increased security.

The link between security apparatuses and light, then, is fundamental. Writing about Jeremy Bentham’s infamous Panopticon – a prison design taking the shape of a central watch tower within a circular rotunda, allowing for constant surveillance (a design much discussed by the likes of Foucault and Deleuze) – Crary highlights the crucial role played by “illumination in Bentham’s original model”. Simply, the panopticon does not only rely on the guard’s physical vantage point to create an ever-present surveyor, but “called for flooding the space with light to eliminate shadows, and to make a condition of full observability synonymous with effects of control.” (ibid). Evidently, for a guard to be able to see anything in the cells that surround his tower, the building must let in a lot of light.

This is no less the case for an algorithmic surveyor than for Bentham’s guard. Let us take the example of facial recognition – for an algorithm to be successful in identifying an individual based on their facial features, his or her face must be well-lit. Any lingering shadow raises the risk of misidentification, and insufficient lighting, overall, is likely to result in a complete inability to recognize a face as such. Ultimately, misidentification is seen as an unfortunate error, a flaw to be corrected. Yet, in an age where the productivity imperative trumps all else, law enforcement increasingly operates under the yoke of quotas, pressuring officers to produce a certain number of arrests or fines, which might otherwise be unwarranted (Ossei-Owusu, 2021). Consequently, an inability to identify faces at all (even if it’s the wrong one) becomes a truly fatal flaw.

The ruthless amassing of data and content for the training of ever-newer and ever-improving algorithms sits at the very centre of the billowing artificial intelligence bubble, which only further promotes productivity at all costs, while generally lowering the quality of what is being produced.

As productivity imperatives have long had an impact on the law enforcement sector, how might this change with the introduction of data-hungry algorithms? Or will the drive to fulfil quotas stay precisely the same? And who will be the most impacted by it?

My latest artwork asks some of the previous questions. Dasha’s Kitchen: My Magical Grilled Cheese Sandwich Recipe delves into the world of algorithmic surveillance through a humorous lens, using review-based YouTube videos. The video is divided into four parts, featuring three distinct characters. The first – Dasha – is a mom, a chef and a YouTube personality, famous for her simple recipes. In this video, she shows us how to make a grilled cheese sandwich, one that will result in a “maaaagical cheese pull.” Halfway through, her video is interrupted by a Russian cooking influencer who starts to criticize Dasha’s recipe, all the while subtly including stories about algorithms, mentioning that her friend who works in IT had recently told her that AI and algorithms are just like a recipe for a grilled cheese sandwich – for the simple reason that both require you to follow a set of simple steps to reach a specific goal. Following a paid advertisement from Thales, the third character (a YouTuber specializing in IT) steps in by trying to explain where the metaphor of a recipe for a grilled cheese sandwich comes from. As it turns out, the metaphor is a common one and can often be found on YouTube to explain the functioning of an algorithm. The third speaker illustrates this phenomenon with preexisting videos that I had found online. The speaker then proceeds to explain how an algorithm can be implemented in real life, using the example of recent experimentation with algorithmic video surveillance by the French state. At the end of this video, the speaker signs off by promoting her course on YOLO and Mediapipe – two open-source machine learning pieces of software which are often used as the backbone of the more recent surveillance and identification technology. So, finally, we see Dasha – the famous chef – reappearing on our screens, and we might wonder, “did we already see a full loop of the video?”, as her environment, outfit, and even her hair look pretty much identical to the first clip. And yet, this is not a loop, similar to cartoon characters and many YouTubers, she has a recognizable look, which only sees minor changes or adjustments from one video to the next, making it impossible to discern the place and time of filming, thereby producing a feeling of déjà vu, which photographer and art historian Julian Stallabrass argues is the result of both the effects of capitalism on the cultural production and the spread of various forms of mechanically reproduced mass media (Stallabrass, 2024). Borrowing from the writings of Vilém Flusser, Stallabrass thinks that artificial intelligence is only further expanding this sense of déjà vu by “[reducing] the complexity of cultural messages, so that images will always show the same thing, and an eternal, endless boredom will spread over society.” (ibid)

So why is Dasha back on our screens? She tells us she has heard about the recent controversies and debates surrounding AI and grilled cheese sandwiches. As her die-hard fans already know, she holds a degree in Computer Science from MIT, so she came to set the record straight. However, it turns out that even our IT specialist remained quite vague, and worse still, promoted the same pieces of software that are built into the tools she was supposedly criticizing. But what exactly does Dasha clearly state with her final remarks? Well, you might have to watch the video to find out.

You can watch Dasha’s Kitchen: My Magical Grilled Cheese Sandwich here.

References

CRARY, Jonathan (2013). 24/7: Late Capitalism and the Ends of Sleep. Verso.

OSSEI -OWUSU, Shaun (2021, May 13). “Police Quotas – NYU Law Review”. NYU Law Review, vol. 96, no. 2 [online]. Available at: https://nyulawreview.org/issues/volume-96-number-2/police-quotas/

STALLABRASS, Julian (2024). “Memories of the Present. Photography and Artificial Intelligence”. New Left Review, no. 148.

Recommended citation: ILINA, Dasha. Who is watching over the city of lights? Mosaic [online], June 2025, no. 204. ISSN: 1696-3296. DOI: https://doi.org/10.7238/m.n204.2507