A security system uses an AI to recognize faces in camera images.
It is supposed to automatically report whenever a person appears.
In many situations the system works very reliably.
It often detects visitors faster than human observers.
But then strange false alarms start to happen:
The AI suddenly reports a face on a house wall, in a cloud, or on a pillow.
In reality, of course, there was no person there.
The software was supposed to detect humans.
So why does it suddenly see faces everywhere?
Question:
Why can an image AI recognize patterns as faces even when no people are there?
Solution will be unlocked tomorrow.