Stuff like that doesn’t always work though, at least on free versions in my experience. I use Ai to write flowery emails to people to sound nice when I normally wouldn’t bother and I used it to negotiate buying my car. I would continually tell it not to use - dashes while writing emails. And inevitably after 1 answer it would go back to using them.
Maybe paid versions are different but on free ones you have to continually correct it.
Even the paid models I’ve tried do that. The style LLMs use seems deeply ingrained. Either companies do it on purpose, or it’s just the result of all the companies using similar training data and techniques.
And it’s the same reason this guy’s solution to not sound like an LLM isn’t a real solution, because no amount of prompting can fix the inherent issue that they don’t actually know anything.
Stuff like that doesn’t always work though, at least on free versions in my experience. I use Ai to write flowery emails to people to sound nice when I normally wouldn’t bother and I used it to negotiate buying my car. I would continually tell it not to use - dashes while writing emails. And inevitably after 1 answer it would go back to using them.
Maybe paid versions are different but on free ones you have to continually correct it.
Even the paid models I’ve tried do that. The style LLMs use seems deeply ingrained. Either companies do it on purpose, or it’s just the result of all the companies using similar training data and techniques.
And it’s the same reason this guy’s solution to not sound like an LLM isn’t a real solution, because no amount of prompting can fix the inherent issue that they don’t actually know anything.