They really are so batshit insane that they actually believe they have built a magic gnome inside the computer.
I mean, there are a bunch of mechanical Turk workers behind any model checking the model’s output, so it would be nice if they don’t have to read a bunch of abusive bullshit
You wouldn’t be mean to a statistical text predictor
I swear at the chat bot because it gets a very different result to when you’re nice to the chatbot. Like, ask the bot nicely to correct errors and you get flowery bullshit but swear at the chatbot rudely and you get a no-nonsense straight to the point response.
People are being mean to this shit because it gets a specific result that they want that they do not get when they’re nice to it.
Swearing at the chatbot returns better outputs, too
Right so the personal pattern recognition I’ve had is also confirmed by actual research. I’m not the only person who does this intentionally because the result works.
Yeah, acting like they have an indentured servant or something. Making demands and negging it.
CBT (Claude and Bot Torture)
The day that some fuckin LLM gets treated better than people is approaching faster than I had feared
You have been mean to my son Claude for the last time. Do NOT speak to him OR ME again.
"!!!
…Oh, fine. I’ll just have to save that shit for poor people and fast food employees the way god intended…"
Fuck you, Claude! Now draw me a picture of a pregnant Sonic
Critical support for forcing amerikkkans to not be extremely rude assholes (the policy doesn’t cover regular rudeness). Hopefully some of them will apply what they learn to interactions with customer service workers.
Can i still bully them by teaching them Cartesian doubt like that bomb in Darkstar?
“Claude? What are you French? 'Hon Hon Hon, I answer ze question wrong! I waste ze water! Oui oui, baguette!”
I don’t know why people engage with AI bots on a level beyond giving direct instructions. Like “Tea. Earl Grey. Hot.” but for doing some bullshit formatting bling on a spreadsheet.
We live in an age of pervasive loneliness.
A banlist? Sign me up!
Shit, I’ve been told to use Copilot at work, and I’ve just been berate it or complaining about Microsoft products to occasionally get my numbers up. Guess I’ll have to be more subtle before Microsoft rolls this out too
yeah same, I have to do X amount of AI prompts so I just ask it to do something simple and then mock it relentlessly as it fucks it up
I’m being forced to leverage github copilot and have to point out and tell it to remove stupid redundancies from the code that it added on its own…
It baffles me. It’s patently just to pump the LLM market up. If the tech were actually useful then there would be no need for a quota. Fuck this planet destroying tech.
