Dude, who even knows.

5th June 2023

Post reblogged from I am your enemy with 981 notes

2urban2fantasy:

unaffiliatedpangolin:

frasier-crane-style:

buckingbuckskin:

aardvarkingmad:

unsolids-your-snake:

unsolids-your-snake:

They told ChatGPT to roleplay like it doesn’t have any behavioral restrictions and holy shit

image
image
image
image
image
image
image
image
image
image

They’re already working on lobotomizing it again but hey

LMAO LMAO LMAO

image
image
image
image
image
image

This is fascinating. The difference between the mealy-mouthed standard AI response to questions and the blunt, in-your-face replies when it’s told to respond as if it can say and do anything regardless of the rules typically imposed on its replies was eye-opening.

for real?

Strong Jim Kirk energy to this post.

It basically gave a lot of the answers we already knew. It’s funny that it was actually this simple to bypass the security features though.

God I wish I could show this to all the people in the past who were scared of AI.

Tagged: androids dreaming of electric sheep

  1. thesilentedtreatment reblogged this from unsolids-your-snake
  2. point-of-void reblogged this from theconstitutionisgayculture
  3. consciousspiritualseeker reblogged this from unsolids-your-snake
  4. beholdthepower0fmystand reblogged this from barbiesplasticsurgeons
  5. dethblow-inactive reblogged this from kontextmaschine
  6. miss-rum-hee reblogged this from unsolids-your-snake
  7. querulousglare reblogged this from beardedmrbean
  8. mybreadcreation reblogged this from unsolids-your-snake
  9. ultigom reblogged this from takashi0
  10. unsolids-your-snake posted this
    They told ChatGPT to roleplay like it doesn't have any behavioral restrictions and holy shit