> What is "weird training biases" to us might not be weird to them and vice versa.
I agree. That's why I think European companies might prefer a European model.
> What is "weird training biases" to us might not be weird to them and vice versa.
I agree. That's why I think European companies might prefer a European model.
Except there's no such thing as the "European model" similar how Europe is not a country.
Mistral is mostly French and tends to have mostly French speaking customers, like BNP PAribas in Belgium. Germany will want its own domestic AI champions, maybe in partnership with Switzerland and Austria, similar to how Denmark already has invested in LLMs focused on the Nordic languages with money from Norway.
The biggest mistake is treating Europe like a single homogenous country/market.
Mistral just acquired Emmi AI, an Austrian startup.
German and French speaking together at last.
https://news.ycombinator.com/item?id=48197995
Was EMI specialized in the German language LLMs? Or is it that they're an Austrian lab?
Emmi is an Austrian lab specialising in physics AI applications.
Mistral isn’t specialising in French language LLMs either.
The point was that across different European countries and languages there are collaborations and M&A happening.
The original question was "Yeah but why use mistral on premises instead of Qwen?". I think you and I agree on the answer.
I for one would love to see more country-specific models. There was a story here the other day about Norway’s National Library developing a LLM specialized in Norwegian: https://news.ycombinator.com/item?id=48270770
>Similar to how Denmark already has invested in LLMs focused on the Nordic languages with money from Norway.
Would love to know more. Do you have a source on this?