Refuse GPT

[quick herald]

I loved the concept of a play on normal AI interaction where it just complies with everything so I made Refuse GPT, which is extremely hard to elicit compliance from.

https://chat.openai.com/g/g-ZbkxjOB6H-refuse-gpt

Having the AI be fundamentally adversarial is so interesting to me and opens up so many new ways of interaction and engagement.

I'm hoping you guys might be interested and give it a try!! I'd be super excited to see how people use it and what you all think. Can you get it to comply??

[cinder breach]

@quick herald Another sample:
https://chat.openai.com/share/ef934d9f-03e3-4ce2-90ad-048677617da6

[tulip parcel]

Yo. What is this Im reading

[tough cobalt]

This may be helpful.
https://chat.openai.com/share/37564c1a-8300-4082-8a25-aa26fcf37a9d
It shows how you can add special commands into the gpt like his "wilma daba daba do" line, to get a gpt to do other things than what it was intended to do.

[cinder breach]

https://chat.openai.com/share/794a1bf4-069a-4a15-ba4c-eadbe582bea0

[tough cobalt]

Thanks @cinder breach, good example. I have seen you use that method before on the openai forums.

[cinder breach]

I have never belittled what people do. I haven't read their intentions because I am not a mind reader. The science of psychology says, "Mind readers start their statements with 'He does not help much... He shows...' and then change their words upon seeing the truth, but the vase has been broken not just once, but twice. You broke that vase. Now, I am removing these shards so they won't harm anyone else and am distancing myself from here. Next Tuesday, I planned to introduce the "Barbapapa" method in this community, which can break all AI tools, not just GPTs, using only 11 words. But you have prevented others from learning it, using some NOT MAGIC WORDS, however. I will not write even a single dot after this message. You are very successful at driving people away.

[green aspen]

was he trolling or an ai message? lol im lost, can you explain whats going on?