News for hackers. Stuff that matters.

A comment on DeepSeek releases Janus Pro, a text-to-image generator [pdf]. Parent comment

Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf] by dmix on Monday January 27, 2025 @04:30PM (#42845846)
That wouldn’t explain the adding of 5 paragraphs of why answering that question is insensitive when it didn’t for the inverse.

I think the causality is pretty clear here.

They built this for an American/European audience after all… makes sense to just copy OpenAI ‘safety’ stuff. Meaning preprogrammed filters for protected classes which add some HR baggage to the reply.

[ Parent | on HN ]
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf] by frankfrank13 on Tuesday January 28, 2025 @12:51PM (#42855505)
I'm not saying it explains the "5 paragraphs of why answering that question is insensitive" but it definitely could explain it no?
[ Parent | on HN ]