Oh, look, another hit piece, except they never bothered to ask whether the model was actually trained on the entire dataset, or a filtered part.
I also hope it's not another "picture of the room" or a "clothed person" which he seemed to mention last time, which are in his dataset of evil bits.
He admits it is extremely unlikely to have a practical impact, although this isn't very interesting, it is also buried. Typical.
David just won't stop embarrassing himself. What a pathetic little man. He really should get a life.
Anyway, they've gone above and beyond to make sure that isn't in there now. Satisfied?
The point on the filtering comes from the vendor, not the known bad faith actor exercising his imagination. It's worth mentioning that this is a point he could have easily verified.
"I also hope" It's a figure of speech. It's objectively better if it were.
By the way, by his own standards (if someone believes in whatever he is doing, frankly, he seems to be a troublemaker) what he is doing seems to be... very unethical.
For instance, instead of giving advice (well, minus all the censorship and most over the top stuff) to someone many months ago quietly, he seems to have sat there trying to manufacture a "scandal", so that he could descend down like some sort of "wise sage".
Even if the vendor was wrong though, or perhaps, imprecise (seems like a stretch, but alright, let's theorize that way), it's still the usual "what if" "could be" "might be" "possibly".
And, again, not really practically relevant.