would humans not make the same mistake?
We also have intuition. Where Something just seems fishy.
Not saying AI can’t handle that. But I assure you that a human would’ve identified a moving cardboard box as suspicious without being told it’s suspicious.
It sounds like this AI was trained more on a whitelist “here are all the possibilities of what marines look like when moving” rather than a black list which is way harder “here are all the things that aren’t suspicious, like what should be an inanimate object changing locations”
But it seems like all these ML models are great at image recognition but not behavior recognition.
What’s the state of the art with that currently?