Hacker News new | ask | show | jobs
by musicale 427 days ago
> reading two documents side by side while looking for errors is much more draining than just reading the original letter and typing it in

Validating LLM-generated text seems to be a hard problem, because it requires a human-quality reader.