5 answers

★ Best answer

You're right to notice the difference 😄 Google completely switched things up around 2016-2017, moving from statistical translation (like rules plus dictionaries) to neural machine translation, so yeah it's AI but not exactly like ChatGPT - it's neural networks trained specifically for translation rather than a general language model. Because of that it understands context and nuances way better instead of just stringing words together one after another, which is why your emails get translated way less weirdly than they used to.

Exactly, it's been neural machine learning since like 2016-2017, not just an upgrade of the old system. The old version used grammatical rules + dictionaries, so it would fail on nuanced contexts and idiomatic expressions. Now the model learns from millions of pairs of translated texts, which lets it understand meaning rather than just matching word for word. That's why your emails suddenly sound less robotic and more natural than before.

The old Google Translate before 2016 really just had dictionaries and rigid patterns, so it was often hilarious with complex sentences. What the others mentioned is spot on, the shift to neural machine learning was radical, but what people tend to forget is that Google keeps refining it constantly - every year they add languages or improve the contextual understanding layer, so it's not like a one-time change in 2016-2017 and then done. It's actually pretty different from ChatGPT, more specialized for translation tasks, so less capable of going off the rails outside its scope but more reliable in its domain.

It's true there was a big change, but it's not really machine learning "like ChatGPT" - it's more neural machine translation, which is different. The old system was actually based on rigid rules and dictionaries, but since 2016-2017 Google has been using neural networks that learn patterns from existing translated texts, so it understands context and nuances much better.

The major difference is that now it can handle complex sentences, idiomatic expressions and keep the tone, instead of translating word-by-word like before. That's why you find it way better on your emails 😊

What language do you usually use for your translations? Because yeah, the shift really happened around 2016 with Neural Machine Translation - it's not exactly like ChatGPT, but it's definitely deep learning instead of the old rigid rules from before.

One thing people don't mention that often: if you check "detect language" instead of forcing it yourself, it usually works better on mixed texts. I also use the post-translation review option where you can mark translations as wrong - that feeds their algorithms directly, so it keeps improving over time.

Sebastien66 asker Mostly English, sometimes German. I'll test the auto-detect and the review option, thanks for the tip!

Your answer

Log into answer.