An interesting article on the possibilities of a non-trivial perception machine, and their history.
At the end, a block on TRM.
Let us cite:

“The Recognition Machine by artists Antje Van Wichelen and SICV (Michael
Murtaugh and Nicolas Malevé) can be seen as one possible enactment of such
a non-trivial perception machine that is also actively counter-imperialist (figs.
3 and 4). I had an opportunity to interact with the version of the work
presented at the Photoszene Cologne festival in May 2019, but the project also
has an online counterpart. Looking like a photo booth, The Recognition Machine
invites gallery visitors to enter and take a digital photo of themselves. The act
of taking a photo activates an algorithm that attempts “to establish links
between the pixels just recorded and those of images from a database of 19th
century anthropometric photographs,” which have been transformed by
analogue techniques. “The resulting print output links contemporary regimes
of surveillance to those of a colonial past.” (“The Recognition Machine”) [9]

The link between the images pivots around the emotions identified by the
algorithm in the viewer’s face and linked with the emotions read in the archival
photos. The reading was obtained by training the algorithm on the FER-2013
dataset, in which each image had been assigned one of seven emotions: anger,
disgust, fear, happiness, sadness, surprise, or neutral. Any possible
misrecognition of emotions that occurs as part of the process serves as an alert
to the system of consequences that predictive technology is imbricated with:
while labelling here is just an innocent game for art audiences, the
misrecognition of image links, their wrong categorisation and ascription, has
serious consequences for the lived lives of many. The visitor may keep the
print obtained, but they are also asked to explore further the posited analogy
and thus go deeper both into the archive and the colonial history of portraiture.

The Recognition Machine also shows us that all images exist as part of the imperial colonial network of visuality, a network that renders some bodies as visible
and proper while deeming others as illegible and/or illegal. What is interesting
about this project is that the artists dispense with the idea of a singular image
as a stand-alone artefact to be admired, classified and otherwise exploited,
showing that all images are part of multiple networks of knowledge and data
exchange. The Recognition Machine thus offers a model of the perception
machine as an invitation to study the production of visuality, the image
networks and their infrastructures, their underlying data and databases, the
algorithms that shape both their production and their networking.”

in Transformations 36, 2022