3 ms·
The point is they're saying completely different: they claim they look at context, and they claim that they can handle use/mention distinctions. State of the ar
by perihelions 5y ago
The point is they're saying completely different: they claim they look at context, and they claim that they can handle use/mention distinctions. State of the art ML can't do that. They're soliciting buy-in from the general public, with assurances that they know are false.
- true_religion 5y agoThe distinction of context could be drawn via human moderation. I think they will set a flag that states if a channel is news. News channels get to report on questionable material without triggering ML, or at least get a human review at some point. Non-news channels would get regular ML.