As much as the default LLMisms are annoying me, it's also a honeymoon period right now where you can even suspect whether something is AI generated based on the default LLM-isms. Word about how to fix their tone has been getting around in academia for a while amongst students trying to pass detection filters, once they're out into the world we can expect to have even more AI generated content masked behind individualized, unique style prompts that aren't immediately recognizable as the default LLM voice.
Pretty easy to condition a prompt with regional idioms and spellyngs.
As much as the default LLMisms are annoying me, it's also a honeymoon period right now where you can even suspect whether something is AI generated based on the default LLM-isms. Word about how to fix their tone has been getting around in academia for a while amongst students trying to pass detection filters, once they're out into the world we can expect to have even more AI generated content masked behind individualized, unique style prompts that aren't immediately recognizable as the default LLM voice.
I've instructed Gemini to never give me the american spellings of words, and to always pronounce Z the proper way. Never give up the fight.