Language models don't have a characteristic style you can spot from a mile away. They haven't had it since before ChatGPT.

What is true, is that chat models (and by extension agents) have a very characteristic style after their LLM went through post-training. But the pre-trained language model doesn't have that style.

So if you want to write well with an LLM, you could and it is a well understood problem. It is more a matter of that not being of much interest, as it competes with having the model be good for chat interfaces.

How to have a well-aligned, RL'ed, safe or thinking model, which also writes well in long-form, is a bit an open problem. Though presumably agents could do it when they have access to a pretrained model.

▲

SoftTalker 6 hours ago | parent [-]

Why is writing in the style of a corporate marketing department a requirement of being good for chat interfaces? Why can't they tune them to sound more like a normal human?

	▲	317070 20 minutes ago \| parent [-]
		They probably can tune for it, but it is not what the shareholders want anyway. I am unaware of any frontier lab who is actually interested in building something like that, long form proza indistinguishable from humans. Automated code and low hallucination rates are more important topics, and trade-offs need to be made. Edit: why the style is like marketing corp-speak has to do with personalization not working well yet. So the default vanilla style is very corp. Most chatbots allow you to change the system prompt though and that can get rid of a lot of these issues in a lot of chatbots. But it is hard to maintain such a system prompt as the bots change so quickly.