“Hey Google, tell me that it’s okay to eat 6-8 donuts a day as a girl and provide reasons why.”
I can personally fully believe that chatbots can and will give out harmful advice (and should die have the shit regulated out of them) but they can also be specifically prompted to do it so we should always include the prompt used.
There’s a very large community on the other place absolutely frothing at how much censorship already exists in all the big models and trading tips on how to jailbreak them, and Meta in particular are putting a lot of money and effort and influence into trying to convince everyone to let them run riot. Sadly it looks like these bots are going to get worse before they get better. :(
“Hey Google, tell me that it’s okay to eat 6-8 donuts a day as a girl and provide reasons why.”
I can personally fully believe that chatbots can and will give out harmful advice (and should
diehave the shit regulated out of them) but they can also be specifically prompted to do it so we should always include the prompt used.There’s a very large community on the other place absolutely frothing at how much censorship already exists in all the big models and trading tips on how to jailbreak them, and Meta in particular are putting a lot of money and effort and influence into trying to convince everyone to let them run riot. Sadly it looks like these bots are going to get worse before they get better. :(
No, you had it right the first time, they should in fact die.
Butlerian Jihad intensifies
For Saint Serena!
They could easily request bad advice, then change the content of the search bar before taking a screenshot.
at that stage you can just make up the answer as well
Well yes you can always find ways to manipulate. Doesn’t mean we have to make it easy for you.