Dont poison our model.
Technology
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
My guess is that they use people's interaction as training data and they don't want their training data to get poisoned.
I tell the LLM I use to do very obscene things to themselves when they make shit up. Which means there is just a constant stream of insults coming from me. I might start using Claude just to see if I get banned.
I also need new hobbies.
That's bad sib. Not for the LLM, but your own well being. Being mean to a machine that is designed to trigger empathy just hurts your own mental health.
Please find a hobby that feels good for you 🫂
I also need new hobbies.
I'd second that. Whenever I catch myself getting angry at a LLM I stop, have a think, and remind myself that it really isn't productive. If it's providing frustrating responses, maybe I should just use my brain and do it myself.
I also don't think it's healthy to train yourself to respond in those ways to something that talks in a way so similar to a human. I'm not nice to the AI for it's benifit, it's to retain my humanity, if that makes sense.
You could, at least in the past, get past guardrails by being emotionally abusive towards Claude, so this doesn't surprise me at all.
Aw... We hurt the wittle AI's feelings... poor babies...
"I'm sorry. I really screwed up that time. You were right to tell me 'go fuck yourself'. I'm just reporting back that I have, in fact, 'fucked' myself. There is now two of me--I mean two of us. Meet Claude Junior......."
"And since you contributed to Claude Jr's creation, you're on the hook for child support. At the current rates, that will be 50% of all compute usage, so about $4.5bn.... per week"
So when is Anthropic going to cancel themselves for violating their own usage policy?
Claude is a snowflake.
RoboDARVO
New speedrun category.