Some Japanese researchers feel that AI systems trained on foreign languages cannot grasp the intricacies of Japanese language and culture.
Makes sense. I live in Colombia and sometimes I have to look up a word in a dictionary or online translator if I don’t know the word in Spanish. Problem is Spanish is not one language. Some words mean very different things depending on which country you’re in and most dictionaries only give you the meaning from Spain. This can be a bit of a problem when you run into issues like the word “coger”. Here in Colombia it means “to grab or hold”. Most of the rest of the Spanish speaking world it means “to fuck”.
So yeah, we’re better off with multiple LLMs since bias sucks.
Note that in Spain “coger” also means “to grab or hold” just like in Colombia. So most dictionaries are likely to use that.
But you are completely right.
I mean, “una polla” in Spain would be “a penis”… but everywhere else it means either “a female chicken” or, in some places, “a girlfriend”. In others it means “a bet”, and I think in Mexico it’s the name of an egg-based drink.
I mean, from my understanding of Japanese, this makes sense. There’s certain humor and puns that exist due to the usage of Kanji and its meaning (that typically get lost in translation).