This post is part of the Collaborative Discussion 2 assignment for the Machine Learning module.
This post reflects on my initial post, peer responses, feedback, and summary post for the second collaborative discussion on AI writers, human oversight and responsible use.
Wall's article made me think about facial recognition less as a model-accuracy problem and more as a deployment problem: what happens when an organisation treats a CNN score as objective evidence? A convolutional neural network can return a probability or similarity score, but it does not produce truth. In a photo app, a wrong match is irritating; in policing, border control, or military use, it can expose someone to surveillance, exclusion, arrest, or harm (Wall, 2019).