Playing around with LZLV-70b 4QM, i am having a great time with the long form responses. However after a while now i am beginning to notice “AI styled writing” I tried pumping up the temperature to 1.5, repetition penalty to 1.3 and even tried mirostat mode 1,2 on the kobold.cpp
If the repetition penalty gets too high, the AI gets nonsensical. If the temperature gets too high, the AI starts to blurt out nonsense. While I am not aiming to have the AI write exactly like a person and be undetectable, after reading it so much it’s gotten pretty repetitive with it’s word usage and sentences.
I’ve tried prompting it to
" Write in a simple easy to read way make your response more conversational and less formal. Aim to increase perplexity by varying your word choices and avoiding predictable phrases "
but it’s…limited.
any tips or methods to make the AI not be so…dry?
From my experience, give it a lot, I mean really a lot, of consistent examples in the prompt how you wish it to respond in different kinds of situations, including the tones and styles. It would work. Try to give it at least 10 to 20 examples for each scenario. Consistency is the key.
You know, I’ve experienced this a lot too.
The short answer for me is, at least for now, the AI is pretty good at understanding what you meant and going along with it (sometimes, and depending on the model), but although it will give you a range of responses, most of them are going to be garbage unless you find specific prompts that output something that YOU want!
From there, once you’ve found some good prompts and a model that give you a style you like, the creativity really has to come from yourself. I know that’s not a great answer, but that’s where we are right now. It can mimic some human seeming responses, but it needs you to feed it the creative and fun off the wall prompts that get it there.
Same thing with stable diffusion: turning the adherence to the prompt way down will get you “creative” solutions…in that they will be different. But different in an unhinged way, not an interesting way.
Alternatively, finding the keywords and model (usually from someone else’s prompt you liked) will get you the right “essence” of what you are looking for, and from there it can feel pretty magical to throw in your own unique prompt and see how the two get combined!