Ask HN: How Do Systems Detect LLM-Written Text? — 44 Points, 61 Comments Debate Detection Methods

Hacker News·low signal

An Ask HN thread on detecting LLM-generated text drew 61 comments on just 44 points (1.39 comment-to-point ratio), indicating deep practitioner engagement. Discussion covered statistical perplexity analysis, watermarking schemes, stylometric fingerprinting, and the fundamental arms-race dynamic where better detectors train better generators. Multiple commenters argued that reliable detection is mathematically impossible as models improve, making the social/policy problem harder than the technical one.