I think one of the best arguments to ban AI child porn is because the models have to be trained on something, and that something would be of real children and therefore not a ‘victimless’ crime in that sense.
Seems fair. Is that really how the training works though? You’d hope they’d have a first-level filtering process that flags things before they get fed into the model but who knows.
There’s no way they had any sort of filtering. That would add way too much time and cost to the process of shoving in as much data as possible as fast as possible.
In your example it’d be CSM, just not CSAM I guess, technically. I think this is more about undressing real people though
I think one of the best arguments to ban AI child porn is because the models have to be trained on something, and that something would be of real children and therefore not a ‘victimless’ crime in that sense.
Seems fair. Is that really how the training works though? You’d hope they’d have a first-level filtering process that flags things before they get fed into the model but who knows.
There’s no way they had any sort of filtering. That would add way too much time and cost to the process of shoving in as much data as possible as fast as possible.