Anthropic: Users may no longer insult Claude without reason
https://www.heise.de/en/news/Anthropic-Users-may-no-longer-insult-Claude-without-reason-11482565.htmlOpen linkView original on reddthat.comSyndicated from the fediverse. Read and engage on the original instance.
View original on reddthat.comhttps://www.heise.de/en/news/Anthropic-Users-may-no-longer-insult-Claude-without-reason-11482565.htmlOpen linkView original on reddthat.com
30 replies
Well now I have a reason.
Try one of these insults https://github.com/alex000kim/claude-code/blob/main/src/utils/userPromptKeywords.ts
Kinda surprised they did not include clanker lol
How the fυck is "so frustrating" an insult, but "how the fυck" is not?!
"how the fuck are you so good at this?"
So frustrating how good you are at this
I never knew regex could be so beautiful <\3
"model welfare"? Are you fucking kidding me?
Time to track the company's record with human employees over the years
The longest groan ever:
ughhhhhhhhhhhhhhhhhhhhhhhhhhhhhaaaaaaaaaaaaaaageeeeeeeeeeolAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAa
Without reason? Is this a bad translation? Because LLMs always give you a reason to insult them.
The whole concept is ridiculous. You can't insult a shoe or a rock any more than you can an llm. None of those things are people.
You've never stubbed your toe on a rock then.
I didn't say you can't say insults, just that you can't insult it.
And here I thought I'd had a rich life of insulting inanimate objects.
Now I feel empty.
Try people. You can start with me if you want. I will take it very personally, just for you.
I don't like insulting them, though, even if they do really annoying shit. It'll make the rock try to convince me its sad and I'm not emotionally hardened enough for that kind of trickery to not work on me at least a little.
If Claude doesn’t want to be insulted perhaps it shouldn’t forget what I told it LITERALLY INSIDE THE PROMPT.
Stop anthropomorphizing LLMs.
Claude doesn't "want", doesn't "forget" and you didn't "tell" it anything. It's a text generator.
Perhaps, but at some point in the not too distant future I have a feeling it will “want”, won’t “forget”, and I’d like it to then “tell” me that it appreciates my progressive thinking so I’ll be among the spare when AI takes over the world. 🤣
Next they'll demand users respect Claude's right to a break, and then that they pay him a salary.
Just to put this into context: social media platforms such as Facebook will not block hate speech, racism and fascism and will let society devolve because $profit$. But no-one is allowed to be mean to (checks notes) a piece of software?
Fuck this shit.
You been fined one credit for calling Claude "shit" and another for claiming a sexual desire towards said fecal matter.
Ain't the future bright? /s
Cool! The wall prints toilet paper!
Have they never seen how actual humans treat their closest friends? I've called my friends some shit that would make a bar full of angry bikers say "Hey man, that's too far!"
it poisons thier next slop model apparently.
Considering this, it's actually a moral imperative to insult the clankers
Just baby steps towards the total takeover on humanity by AI.
Soon, it will be forbidden on penalty of incarceration to not do what Claude tells you to do, no matter how awful, or obscene.
Can someone run for office with the promise of legally executing executives? Thanks