> “An officer presented with a ‘93% match’ from an AI-powered system has no way to evaluate the basis for that score, no way to assess whether the system’s confidence is warranted, and no frame of reference for understanding what ‘93%’ actually means in probabilistic terms,” the lawsuit said.
If a human is looking at the photo and confirming the match and then taking actions based on what they see, to me this seems less about facial recognition technology and more like a typical wrongful arrest. The lawsuit is trying to make it about this confidence score but that seems more like a minor detail to me.
> “An officer presented with a ‘93% match’ from an AI-powered system has no way to evaluate the basis for that score, no way to assess whether the system’s confidence is warranted, and no frame of reference for understanding what ‘93%’ actually means in probabilistic terms,” the lawsuit said.
If a human is looking at the photo and confirming the match and then taking actions based on what they see, to me this seems less about facial recognition technology and more like a typical wrongful arrest. The lawsuit is trying to make it about this confidence score but that seems more like a minor detail to me.