Hacker News new | ask | show | jobs
by anaganisk 1051 days ago
What if the AI is designed to flag them as potential bad actors.
1 comments

Flag the person who cannot be photographed? What are we flagging again?
Clearly all such people that cannot be photographed are all the same person
You need to look at it from the perspective of the system. The system has a database of individuals and is asked "does this photo matches any of the individuals in your database?"

If someone whose photo is basically a dark circle with eyes is entered in the system then the next time you ask the system to identify a photo that is basically a dark circle with eyes the system will respond to with the previous match.

Like a bloom filter with a bitvector of size 1.