The article argues that the distinction between “writing with AI” and “writing by AI” is misleading, because using AI to draft, revise, or improve text still transfers core parts of the writing process to the model. It explains that human learning from other writers is fundamentally different from how language models generate text from training data, making common comparisons between the two weak. The main point is that relying on AI for composition changes authorship and creative labor, even when a human edits or curates the final result. 

https://jamescosullivan.substack.com/p/writing-with-ai

Joan Westenberg argues that the current witchhunt against AI-assisted writing is misguided and harmful, highlighting the unreliable nature of AI detection tools and the unfair career damage suffered by authors like Mia Ballard, whose novel was pulled amid unproven AI-use accusations. She compares this to historical collaborative writing practices, emphasizing that attempts to police AI usage through crowdsourced suspicion and flawed detectors threaten all writers without offering meaningful solutions, calling for more reasoned and evidence-based approaches instead.

https://www.joanwestenberg.com/the-ai-writing-witchhunt-is-pointless/

The article argues that with AI and large language models making competent outputs cheap and abundant, the real competitive advantage has shifted to human judgment, or “taste,” which involves distinguishing what is generic from what is meaningful under uncertainty. However, taste alone is not sufficient; humans must combine it with real-world context, stakes, and authorship to build genuinely valuable work beyond AI-generated average outputs.

https://rajnandan.com/posts/taste-in-the-age-of-ai-and-llms/