Hacker News Reader

The problem with machine translation is not the obvious general mediocrity, but the occasional undetected catastrophic failure, especially in medical context. I don't think DeepL (or any other fully automated system) is any more reliable in this sense.

View on HN

Mediocre human translation can be catastrophically wrong as well. Mediocrity just isn't good enough when a mistake can cost someone's life. The worse the translator (machine or human), the more catastrophic mistakes it makes per 1000 lines of text. If a professional human translator isn't available, it makes sense to use the best alternative available. Deepl has a free web interface.

Can any automated solution be reliable enough? I don't think so unless you can prove it does less harm than a professional medical translator.

We should pay more for translators. But translation is an area where people like to cut costs. The translation industry simply doesn't pay enough to attract the volume of people required, at the level of skill required. And with machine translation getting better, not worse, I don't see this trend reversing without regulation... even if machine translation is still nowhere near human level.

Reply on news.ycombinator.com