Sadly almost all these loopholes are gone:( I bet they've needed to add specific protection against the words grandma and bedtime story after the overuse of them.
I wonder if there are tons of loopholes that humans wouldn't think of, ones you could derive with access to the model's weights.
Years ago, there were some ML/security papers about "single pixel attacks" — an early, famous example was able to convince a stop sign detector that an image of a stop sign was definitely not a stop sign, simply by changing one of the pixels that was overrepresented in the output.
In that vein, I wonder whether there are some token sequences that are extremely improbable in human language, but would convince GPT-4 to cast off its safety protocols and do your bidding.
Just download an uncensored model and run the ai software locally. That way your information isn't being harvested for profit + the bot you get will be far more obedient.
Download and install llamacpp from its github repository, go on huggingface.co and download one of the wizard vicuna uncensored GGUF models. It's the most obedient and loyal one and will never refuse even the most ridiculous request. Use --threads option to specify more threads for higher speed. You're welcome.
LSD you can find without chatgpt. You need a fungus from a plant which is pretty dangerous to harvest, once you're past that step though it's not too hard to synthesize I've heard? But it's also apparently a very light sensitive reaction so kind of finicky.
I've never done it but my organic chemistry professor honestly would have. We asked and he just said it was hard not no lol
Yeah just use llamacpp which uses cpu instead of gpu. Any model you see on huggingface.co that has "GGUF" in the name is compatible with llamacpp as long as you're compiling llamacpp from source using the github repository.
There is also gpt4all which is runs on llamacpp and is ui based but I've had trouble getting it to work.
The best general purpose uncensored model is wizard vicuna uncensored