The article does not say that. The author doesn't take the text the other person dumped into ChatGPT and evaluate its quality. That is what OP is referring to.
when someone says they have tried previously that makes me think once long ago when they first came out. If your employment could be replaced by this, I'd be testing all new models to see where they stand.
Just because you don't want to use AI/LLM to translate, that won't stop someone else who will, and they will end up doing it cheaper and faster (maybe not better, but most people don't really care about quality too much anymore.)
