> To avoid collecting it despite the expressed wishes of the user, they'd need to be able to recognize it as untouchable.

> And recognizing the data is the exact problem that this African firm was hired to help with. What do you want Meta to do?

This is written as if logically exhaustive, but it misses the very obvious alternative that none of these videos should have been reviewed by a human at all (aka no reason to "recognize it as untouchable"; they're all untouchable).

If you want to get stricter and talk about collecting at all, Meta already has that solution too, by leaving the video in the user's camera roll. Let the user manually add the video to the Meta AI app or whatever if they want to share it with others there.

▲

thaumasiotes 2 hours ago | parent [-]

> This is written as if logically exhaustive, but it misses the very obvious alternative that none of these videos should have been reviewed by a human at all (aka no reason to "recognize it as untouchable"; they're all untouchable).

No, taking that approach would mean that when someone sends you data that you aren't supposed to collect, you collect it anyway. This is the opposite of what was suggested above.

	▲	magicalist 2 hours ago \| parent [-]
		> No, taking that approach would mean that when someone sends you data that you aren't supposed to collect, you collect it anyway. This is the opposite of what was suggested above. That was in reference to the original story, that human annotation is happening on videos that no one knew were getting reviewed. If you want to talk about not collecting at all, well: > If you want to get stricter and talk about collecting at all, Meta already has that solution too, by leaving the video in the user's camera roll. Let the user manually add the video to the Meta AI app or whatever if they want to share it with others there.