I’m thrilled to announce another publication from our European Research Council (ERC)-funded research project on Machine Vision: Gabriele de Setaand Anya Shchetvina‘s paper analysing how Chinese AI companies visually present machine vision technologies. They find that the Chinese machine vision imaginary is global, blue and competitive. 

De Seta, Gabriele, and Anya Shchetvina. “Imagining Machine Vision: Four Visual Registers from the Chinese AI Industry.” AI & SOCIETY, August 1, 2023. https://doi.org/10.1007/s00146-023-01733-x.

You’ve probably noticed that machine vision technologies like facial recognition and surveillance tend to be blue. A simple google image search for facial recognition makes that pretty clear:

Gabriele and Anya took this initial impression and dug a lot deeper, looking specifically at Chinese websites and how tech companies in China are promoting their machine vision technologies visually. They have a lot of fascinating insights – for instance, although these companies are selling surveillance and image recognition systems they rarely mention the technology itself, and the visual imagery they use is more suggestive than concrete. 

And yes, the colours are mostly blue and grey.

In the paper, Gabriele and Anya identify four visual registers: computational abstraction, human–machine coordination, smooth everyday, and dashboard realism. 

1. Computational abstraction

The first of the four visual registers they identity is computational abstraction. I have a feeling Elon Musk would like this aesthetics. It’s similar to what the google image search I showed you above tends to focus on, with the blues and blacks, but it’s more abstract. Gabriele and Anya describe this style as featuring “fluid animations and frictionless connections between parts,” giving a “computational abstraction” that “supports a modular view of machine vision”.

2. Human-machine coordination

This visual register anthropomorphises the technology, and promotes “inspirational encounters between users and machine vision technologies”. It tends to rely on popular tropes about sentient AI, and often portrays the AI and machine vision as more intelligent than humans.

3. Smooth everyday

This visual register is quite different from the previous, and seems more like a combination of smooth Instagram-style images. Gabriele and Anya write that “With its bright, minimalist and clean aesthetics, this visual register seeks to represent machine vision as effortlessly integrated in everyday human sociality – people go about their life in naturalistic settings, and technologies are either invisible or peripheral parts of the picture.”

The function of machine vision, in these images, is to “smooth” our everyday life. It makes me think of the idea of “domestication of technology“, but it’s almost the opposite – it’s almost as though the technology domesticates us, keeping us safe and smoothing things out. As Gabriele and Anya write:

Machine vision makes everyday life smoother – this visual register reinforces a simple message by depicting a very specific range of people and environments with no room for imperfections, failure or doubt. The use of stock images, choreographed scenes and scripted testimonials results in a sanitized representation of everyday life with an overt normative coding of gender, ethnicity and class. Throughout all the websites we analyzed, men and women are portrayed in markedly different ways: men in professional settings, women in domestic ones; men access buildings or drive vehicles, women shop or enjoy leisure time.

In my upcoming book, Machine Vision: How Algorithms are Changing the Way We See the World, one of the chapters analyses Flock Safety‘s surveillance cameras, and they use a very similar aesthetic to this in their marketing. (Side note: the book is at the printers now, my editor at Polity Press tells me, and it will be available next month!)

4. Dashboard realism

Dashboards are a well established part of the smart city imaginary. While the Smooth Everyday register shows how machine vision can tame individuals and everyday life, the Dashboard Realism register shows how it can control cities and whole populations by making them “legible and controllable” (Gabriele and Anya take the “legible and controllable” quote from an article by Jathan Sadowski on dashboards in urban informatics).

Conclusions

Gabriele and Anya summarise their findings by showing how the four visual registers relate to each other. 

I think this article is a really valuable contribution to the discussion of how visual presentation builds upon and develops our existing sociotechnical imaginaries. In some of our other work in the Machine Vision project we have focused on how machine vision technologies are represented in science fiction, digital art and video games. This paper examines how advertisers use it in the Chinese context. Gabriele and Anya don’t just identity these four visual registers, they provide nuanced analysis showing the effects of these ways of portraying technology.

From our analysis of Chinese corporate websites, we conclude that the sociotechnical imaginary articulated by these four visual registers hides the functioning of machine vision technologies behind the abstract and the fantastic, and depicts their use through narrow representations of everyday life and actual use. 

Another contribution is the method they use, which they call a multimodal clickthrough method. This builds upon the much-used walkthrough methodthat Ben Light, Jean Burgess and Stefanie Duguay developed for analysing apps. After deciding which websites to study, ,Gabriele and Anya’s strategy was “following links, playing videos, interacting with tech demos, exploring API documentation, and occasionally looking at source code”, and collecting both screenshots and descriptive notes. They then used multimodal social semiotics to analyse their material.

This is the final year of the Machine Vision project, and we have a number of publications coming out, so look out for more from us!