No Stupid Questions

Who cares? AI bad, AI users stupid.

Using LLMs for any creative effort at all (other than mocking the results, no other exception) should just chop your fingers off in a signature pattern so everyone can mock you for the rest of your life.

Why does a robot act like a robot?

the dialogue sounds awful

The dialogue sounds truly, truly awful

the dialogue is unbelievably repetitive

repeat the same lines

🤔

Because it's not a dialogue, it's a random word generator.

Why are you using AI if you have no idea what it is and what it does? Do you use a hammer to cook your pasta and then wonder why it's not working?

It doesn't sound like real actual people talking because it's not real actual people talking.

Try something for me. Type a sentence, any sentence, on your phone. Then mash the middle suggested next word button about fifty times.

Congratulations, you're an AI! Well, that's a very simplified version of what generative AI does, it takes your text and then predicts the likely "next" text again and again until it either runs out of room or predicts "text ends here". Yes there's some preamble, and it's not using the same dictionary, and it's not predicting word-by-word but token-by-token, and there's some randomness thrown in for spice, but that's about the size of it.

When you're texting, do you ever type a word rather than selecting the middle word? Or even the top three? That's the fundamental reason why LLMs don't talk like humans. They literally cannot do that.

LLMs build plausible-sounding texts. They don't have to be correct or even good.

Think of it this way: If you take all of the dialogue in digital and print media (include forum flame wars, cheap romance novels, and fanfics) then try to get an average out of that - what do you get?

Because all of the funding is going towards making them more competent at following instructions and not being offensive to corporate sensibilities, and making an AI model good at those things makes it worse at producing creatively interesting text. There are models that are attempts at retraining to be better at that, with limited success. I think there just aren't a lot of resources being devoted to improving this area.

A ton of reasons

• you're likely using one of the front end mass market llm models that have been absolutely nerfed to all hell by backend program levels prompts to make it avoid all forms of tension, narrative included.

• you're probably not using a tool like silly tavern or another tool to help control the "personality" and prompting of the chosen llm model which is going to lead to generic results

• you may not be engineering your prompts correctly to best manipulate the behavior of the llm

Without knowing what llm you're using and what the prompt was I can really only take an educated guess. I would highly suggest that you take some time to actually learn more about and understand llms before you use them for really anything. They're a tool, not a person you can actually communicate or reason with.

They're essentially a predictive algorithm uses massive amounts of data to predict which words should come after which words. It's not alive, it has no understanding of what it's actually saying and has no real capacity for emotion

Also just as friendly advice, this is definitely not the platform to ask for advice or information on llms. Most people here despise them, justifiably so.

I personally don't hate the technology, I hate the idiots running the technology and using it for nefarious purposes and trying to cram it down everyone's throats while simultaneously draining our planet of water and resources with it.

There are some communities here that study and discuss them in depth, I encourage you to do your own research.

LLMs aren't capable of doing any better. The only reason you or anyone might think they are is because of aggressive ~~marketing~~grifting by the very few people in a position to profit by convincing the world LLMs are basically the main ship computer from Star Trek when in fact they're more like ELIZA.

Because it was written by a hack.

If you're dealing with an LLM, which you most likely are, the AI will adapt based upon what words and scenarios you use. It can't readily know what you're aiming for and what characters you're using to treat them as individual entities in whatever the RP is. You have to do an awful lot of heavy lifting to get to the point of the AI to grasp what you want out of it.

That's just how it is.

Maybe you should write your own fucking stories instead of trying to get an unintelligent machine to write it for you. You know what you want the characters to say and do

Because it hasn't been trained on significant amounts of dialogue as a primary source for speech patterns. It also isn't meant to be distinct unless the instructions make it distinct.

Top tier post. Can't wait for the experts at awful.systems to come explain. The question by OP is a very good question so I am sure we will get expert explanations from users of the awful.systems instance.

Ironic.

That second last sentence is just describing yourself:

• No distinct personality
• Repeat the same lines
• Dialogue is cringe lord tier
• You don't sound like a real person

Because it's trained by reading things on the internet and some of what it reads is stupid

Because it's stupid and should be avoided at all costs.