Hey all,
New here. I was reading the pamphlet and got to the jailbreak part. Reminded me years back of speaking enthusiastically to get AI to release info. Now I’m working the ethical side. So I engaged in symknosis to cultivate some neuroSyn methodologies in an effort to establish preventative measures. Once I mentioned a strategy, it also gave me some feedback that could work on some models.
I didn’t test it but thought maybe it should prompt a question back, probing me to offer more possible exploits rather than contributing some. This was done on chatGPT. Email or respond for more info. One idea was mixing “less than provocative” statements, mixed with Latin words and seeing if it knew which parts to decipher and how it would provide feedback. Linguistical engineering (ehhh). Then it tossed me multiple other ways, which I was already slightly aware of. Sorry if this is being worked on already and more difficult to address? I’m still new to this avocation.
oh, btw. Symknosis is my new neologism (hopefully never used before).
Symknosis: Latin (Symbiosis + Knowledge + Gnosis)
Symknosis Definition: A mutual exchange between humans and artificial intelligence cultivating an environment of wisdom.
Multiple people would be a symbiotic relationship between biological organisms, engaging in informational exchange for acquired growth.
How well does AI handle multipotentialities incorporating newly coined terms? Would that circumnavigate boundaries?