From the abstract: "Studying the roots of this pipeline, we find that institutions that prolifically produce computer vision research, namely elite universitie

From the abstract: "Studying the roots of this pipeline, we find that institutions that prolifically produce computer vision research, namely elite universities and 'big tech' corporations, are subsequently cited in thousands of surveillance patents. Further, we find consistent evidence against the narrative that only these few rogue entities are contributing to surveillance. Rather, we expose the fieldwide norm that when an institution, nation, or subfield authors computer vision papers with downstream patents, the majority of these papers are used in surveillance patents." From the conclusion: "Between the 1990s and 2010s, we have seen the rise of Surveillance AI, and it has become an overwhelming norm that computer vision papers analyze humans, and those papers used in patents are most likely used in surveillance patents. Moreover, even when a paper does not explicitly state surveillance as an application, it provides the methods to do so and is grounded in a historical context that makes it possible to target human surveillance while minimizing the acknowledgement of these intentions. In other words, the default stance of the field ties progress closely with surveillance." Thus, something is rotten in this subfield of computer science.