Synthetic Data and Reverse Image Search: Constructing New Surveillant Indexicalities

Abstract

The recent evolution of algorithmic techniques (mining, filtering, modelling) makes people more transparent through sophisticated search interactions and online monitoring, heightening opportunities for surveillance. The advent of computer-generated “synthetic data” has created another twist in the techno-information revolution of generative “artificial intelligence.” Promoted by tech companies to circumvent privacy legislation and to develop cheaper monitoring technologies, synthetic data is touted as a solution to surveillance capitalism. This dialogue paper focuses on the use of synthetic data in the context of “fake” images and discriminatory technologies by first discussing the relation between representation and indexicality via the medium of (digital) photography and then via reverse image search. A digital ethnography by the authors uses artificially generated images of people “who do not exist” to query the reverse search engine PimEyes, which offers a biometric search for anyone wishing to find their faces on the internet. PimEyes finds faces similar to a person that doesn’t exist, provoking questions both about the generated image used as a query and the status of the search result. The results show the tensions inherent to the use of synthetic data: a dialectic between increasing precision and increasing scepticism. When visiting the offered, linked websites, the confusion increases as the user struggles to determine if the images PimEyes found are synthetic or real. In this context, reverse image search will likely stimulate future synthetic data development and simultaneously offer services that embed metadata into files, as well as forensics to secure indexicality, introducing yet other factors into the loop between representation and generation. Therefore, the matter of concern won’t be the ability to produce realistic representations through the use of synthetic data, but the demand of indexicality that their use triggers and the bureaucratic apparatuses of verification that emerge to contain it.