News for hackers. Stuff that matters.
Re:Show HN: Get conversational practice in over 20 languages... by adamsmith143 on Tuesday September 27, 2022 @10:15AM (#32995485)
If you slap an M rating on it why would that matter?
Re:Show HN: Get conversational practice in over 20 languages... by Gigachad on Tuesday September 27, 2022 @07:29PM (#33002477)
Because an NPC will inevitably be made to say something racist which will be screenshot and posted on Twitter.
Re:Show HN: Get conversational practice in over 20 languages... by Loughla on Tuesday September 27, 2022 @10:18AM (#32995518)
I believe no company in the world would want to put out a game where an NPC could be coaxed into presenting a sexy story featuring an infant Adolph Hitler and Jesus Christ.

It's just a headache that I wouldn't want, I guess.

Re:Show HN: Get conversational practice in over 20 languages... by cbozeman on Tuesday September 27, 2022 @10:33AM (#32995740)
Or maybe we just need to evolve and realize that people are going to use tools for what they want them to be used for, not what we want them to be used for.

People have apparently already found workarounds for Midjourney's nudity filter.

StabilityAI was smart enough to realize people will get around that, and allowed the objectionable content filter to be removed via commenting out a few lines of code.

It's our sensibilities that need to change, not our tools.

Re:Show HN: Get conversational practice in over 20 languages... by adamsmith143 on Tuesday September 27, 2022 @10:21AM (#32995550)
There are filters on the most common LLMs so you could obviously tune outputs to be within some range of possible outcomes.