Hacker News Reader

What if my "AI Character" says things that are offensive or insensitive or considered rude by someone in some culture? I can see this making headlines with a <Celeb/Politician> AIs offending someone...

View on HN

4chan is going to have a party with this and there’s no way to implement enough guardrail to prevent all the potential failure modes.

I don’t think this will end well for them.

4chan is not going to have much fun with this because of LLaMA being quite filtered and sanitized. It was character.ai that first popularized the AI character roleplay genre, but then they implemented quite strict filtering, so nowadays there are literally tens of uncensored NSFW platforms for AI character roleplay, and people can easily download local models that just require a good-enough GPU. Or abuse models from OpenAI/Anthropic.

Of course with the magic of browser web tools, the AI character doesn't actually have to say anything offensive, you can just change whatever it did say and post screenshots on Mastodon.

Reply on news.ycombinator.com