Plus, I’m wondering how you do it without data leaving the device. All I can think of is an AI model that willl process every picture. You can’t compare to known hashes, b/c it’s a novel picture out of the camera. The device has to judge it somehow.
If there’s an AI model, you’ll have false positives and false negatives. And gray areas, where even reasonable humans would disagree. Since false negatives are way more legally risky, co’s will err waaaaaaaay on the side of false positives. Which will block totally innocent pics. And gender diffs in what’s socially accepted, and the AI will get that wrong a lot.
Plus, I’m wondering how you do it without data leaving the device. All I can think of is an AI model that willl process every picture. You can’t compare to known hashes, b/c it’s a novel picture out of the camera. The device has to judge it somehow.
If there’s an AI model, you’ll have false positives and false negatives. And gray areas, where even reasonable humans would disagree. Since false negatives are way more legally risky, co’s will err waaaaaaaay on the side of false positives. Which will block totally innocent pics. And gender diffs in what’s socially accepted, and the AI will get that wrong a lot.
Gonna be a mess.