Face Recognition & Its Ethics

Your phone doesn't contain a 'you' class — nobody could train a category per person on Earth. Instead, face systems detect the face, embed it into a vector, and compare distances.

Face unlock is a 1-to-1 check: does this face's vector sit close to the enrolled one? Airport surveillance is 1-to-N: match one vector against millions in a database. The embedding is trained so the same face stays close across lighting, angles, glasses, and years — while different faces are pushed far apart.