Interesting. I wonder if in future, people who use locales that are not supported by ChatGPTs/Ernies would be more disadvantaged. (They are probably already disadvantaged to some extent)
I hope LlaMa/Alpaca come to a level even these people can train it for their own language.
Translation is not enough, you also need locally relevant knowledge. If you ask a language model in Khmer to explain how to file your taxes and it replies in perfect Khmer, cheerfully telling you all about how taxes work in the US, that's not very useful if you're actually in Cambodia.
I hope LlaMa/Alpaca come to a level even these people can train it for their own language.