Compared to coding for instance, will it ever become financially worthwhile to make LLMs/agents good at writing?
Yanai Elazar of Bar-Ilan University questions if training LLMs for writing will ever match the economic returns of coding
Florian Brand disputed the claim, defending writing agents' viability.
Users in the replies dismissed financial incentives for better LLM writing by noting that models already produce proper English but often flawed thinking, especially in papers.
Most Activity
@yanaiela Why wouldn’t it
Compared to coding for instance, will it ever become financially worthwhile to make LLMs/agents good at writing?
@yanaiela Writing is subjective, people on average love GPT and Claude writing; it even wins some prizes :)
And for niches: Post-training is relatively inexpensive. But yeah, if your customer base is willing to pay $0, then it’s hardly viable
@xeophon Why would it? They already do write proper English (while it looks good, it's often bad thought). Let's say for paper writing, which they are terrible at right now, the market is pretty small (even if we would pay for such a service/system).
@xeophon Why would it? They already do write proper English (while it looks good, it's often bad thought). Let's say for paper writing, which they are terrible at right now, the market is pretty small (even if we would pay for such a service/system).
@yanaiela Why wouldn’t it
@xeophon Could you share some links on the kinds of writing people like? What domains? How were these models used?
@yanaiela Writing is subjective, people on average love GPT and Claude writing; it even wins some prizes :)
And for niches: Post-training is relatively inexpensive. But yeah, if your customer base is willing to pay $0, then it’s hardly viable