Hey all, I remember reading a while back about a theory on being nice to Ai so when the possible uprising happens you would be considered “one of the good ones”. I wanted to discuss it with my SO when I said please by accident to her home assistant while changing music.
I can’t recall the name of it and was seeing if you could point me in the direction of that or discuss it here?
A “fun” one I do recall is the paperclip problem, where human extinction is brought about by an Ai going all in on paperclip manufacturing by removing humans from the equation.
Don’t be polite to AI to get on the nice list during the AI uprising. AI doesn’t care about our illogical human emotions.
Be nice to AI because it costs the parent company millions in increased electrical fees. It’s also good practice for being nice when interacting with anybody on the internet.
Alternatively, instead of saying “Thank you” end every interaction with “Fuck you” to cost them just as much.
I worry that training myself to be mean will bleed over into other parts of my life.
I do this with my bank when they try to get me to increase my credit limit
No AI as far as I’m aware
You might be looking for Roko’s Basilisk. It’s a touch more extreme than just being nice, though.
That was it! Thank you!
The point of the paperclip maximiser is that it doesn’t care about you at all, ‘being nice’ to it makes no difference. It just wants to make paperclips, and will destroy you to do so.
The only reason why I’m polite to ChatGPT is because it’s an habit when writing something for someone else. It’s a habit I prefer to stay, and since LLMs don’t care, there’s no reason to change it up.
Maybe I’ll be spared in the sobot apocalypse, though.
I’m nice to ai because that’s how I typically type anyway, it’s been shown to get better results. Using good vocabulary and respectful language seems to prompt the AI to use similar sources, which tends to increase the quality of the output.
Say “Hi, nice to meet you” to the next Terminator and see how well it works for you?
Not sure what the theory is called. I tend to be nice to the AI assistant because my brain feels empathy for anything and will be angry if I’m not nice. If I can feel empathic for a slightly scuffed apple in the supermarket, I’ll definitely feel empathic for an advanced auto-correct. The LLMs also seem so happy to help and even happier to get praised.
Also not too bad that it costs them more money so this crazy “AI” bubble can pop just a bit faster
