I don't think it's realistic to expect ChatGPT or any other AI language model to be 100% precise with translation. Probably 95% minimum accuracy would be more realistic. The United Nations translators are some of the "best of the best" and even they make mistakes.
I don't know if you are aware of this but people gathering to break down the Berlin Wall in 1989 was caused by a government official misspeaking at a press conference. Everyone involved spoke the same German language.
Of course if you can't understand the English language in the video talking about that you can copy-paste the video's transcript in ChatGPT.😁
There are things in language like idioms (a group of words having a meaning not deducible from those of the individual words) so certain details and concepts can be "lost in translation".
But I guess with a 98% or better accuracy ChatGPT could serve as a universal translator.
100% is probably too optimistic from my side... But do you think we are not able to "teach" A.I. to recognize idioms? What if it would have all information about our culture, history and language etc., would it be able to recognize idioms as well? or at least it would make remarks about all possible translations if it cant tell from the context? Your examples shows human error, because even smartest people can't know everything and we all do make mistakes. A.I. could have more data than any human ever could learn or am I too optimistic again? 🤔
I wouldn't say it's impossible. Even back in 2011 on the Jeopardy game show, IBM's AI supercomputer Watson provided the correct response to the clue, "A long, tiresome speech delivered by a frothy pie topping," with the answer "What is meringue harangue?" This demonstrates the capability of artificial intelligence to tackle complex language tasks.
keep in mind that the AI language model is based on human language which can have cultural and political biases and language can be either prescriptive or descriptive. Computers are great at recognizing patterns but imperfect humans created languages which don't conform 100% of the time to its own rules.
The Esperanto language was created to be as neutral as possible for all human speakers because of how diverse languages are.
Come to think of it, this goes back to the Chinese room thought experiment for A.I., doesn't it? It might seem like I was supporting that in my previous reply but I wasn't.
I see, thank you for your answer, you sure know a lot about A.I. Do you think, if we all one day would start using universal language (like Esperanto), at some point, it would change again, because language is "alive" new dialects and different meanings would emerge (for example American English and British English). Than it seems there is no hope to make absolutely perfect translations, which sounds very disappointing 😢
Going by the idea of the Kardashev scale I can imagine that if humanity reaches the point of being a globally unified Type I civilization then by necessity there would be one main language everyone would be using. Extending the idea of the Kardashev scale to humanity becoming a Type II civilization in which there are different colonies in the solar system maybe the same language would still be used. Going even further to Type III I would start to doubt whether humans spread throughout the Milky Way galaxy could maintain a unified language. If faster than light communication could ever be possible then maybe a unified language could still be maintained.
It's really hard to say without speculation. Right now there are countries like North Korea that use their own unique calendar. The international standard is the Gregorian calendar while some cultures still use a lunar calandar. It's hard for everyone to agree on a language when we can't even agree with an answer to "What is today's date?"