Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

EuroLLM: LLM made in Europe built to support all 24 official EU languages

eurollm.io

Post reply on HN
Print view View original

141–150 of 629 posts

Re: EuroLLM: LLM made in Europe built to support all 24 official EU languages

#141

Post by wongarsu »

sorenjan wrote: post #18

If I want to use an LLM to do translation, should I use a base model or an instruction tuned version? I've had mixed results using the chat models and a simple "Translate this to : "

For a 9B model like EuroLLM, fine tuning the base model is pretty viable. You don't need a lot of samples, on the order of 300 high quality examples can produce good results, and the GPU time is pretty manageable with rented GPU instances

Just the base model and a template like "English: {text}\n{language}:" can also work with a bit of filter and retry logic

Re: EuroLLM: LLM made in Europe built to support all 24 official EU languages

#142

Post by miohtama »

KaiserPro wrote: post #113
whimsicalism wrote: post #104

Earlier quoted context omitted.

> When you talk to most EU business owners, even in tech, the limiting factor isn't regulations. This being the #1 reason is such a tired trope. Okay, what is the limiting factor? Because when I talk to EU business owners (admittedly, very few) - they point to lack of big EU capital markets, which is directly downstream of the policy environment. And when I talk to top EU human capital, they all point to the lack of…

Money. Why work in the "europoor" countries when you can go to america and earn megabucks.

That's capital markets and the lack of capital markets is because of not having business friendly environment. Consumer protections strong, pro business not so much. Companies like Spotify go to the US to IPO.

Re: EuroLLM: LLM made in Europe built to support all 24 official EU languages

#143

Post by ffsm8 »

actionfromafar wrote: post #136
whimsicalism wrote: post #124

Earlier quoted context omitted.

Okay, can you please quote me directly what I said that is unhinged/delusional or explain why he is on point that I should seek help? I think you have lost sight of that being an incredibly rude thing to say to someone, but maybe not?

Feudalism is a problem in some corners, but it's mostly a UK thing. The US has its own almost-feudalism like ownership structure (see 1980s to now wealth distribution), which is problematic, but it's not bad enough (yet) to halt the US economy.

And to add to your point: The UK isn't even part of the EU...

Re: EuroLLM: LLM made in Europe built to support all 24 official EU languages

#144

Post by ardit33 »

NicuCalcea wrote: post #74
kreetx wrote: post #33

Earlier quoted context omitted.

Yup, most of Eastern Europe are Balto-Slavic. While the division from the Eastern Slavic languages (Russian, Belarussian, Ukranian, etc) is distant, they are still Slavic. From Eastern Europe, only Estonian is not a Slavic language.

> From Eastern Europe, only Estonian is not a Slavic language. Well, that and Romanian. And Hungarian. And outside the EU, Albanian. And Georgian, Azeri and Armenian if you consider those Eastern Europe.

Albania is not "East Europe", but South East. Same as Greece.

Re: EuroLLM: LLM made in Europe built to support all 24 official EU languages

#145

Post by tinco »

whimsicalism wrote: post #57

[flagged]

Yeah no, it's just not how it works. They're trying to support fundamental research and they have limited resources to accomplish them. Some random dude who wants to build a company that generates pretty AI pictures is just not the target audience, and he rightly got rejected.

And frankly, the dream scenario that Pieter describes where he somehow would qualify for these resources also wouldn't help kickstart the tech industry, and it's also not how it works in the states.

What does help, and what European governments (at least the one in The Netherlands that Pieter is from) actually do, is more funding for startups. If you're a startup founder in NL almost every angel you talk to has a matched funding deal with the government. That's such a smart way of keeping up with the US. Do you think US startups get free compute from the government? They don't even get subsidies most of the time. What they get is better funding because there's more capital available, and helping investors with that is exactly how you solve that.

Re: EuroLLM: LLM made in Europe built to support all 24 official EU languages

#146

Post by ardit33 »

rich_sasha wrote: post #94
kreetx wrote: post #33

Earlier quoted context omitted.

Yup, most of Eastern Europe are Balto-Slavic. While the division from the Eastern Slavic languages (Russian, Belarussian, Ukranian, etc) is distant, they are still Slavic. From Eastern Europe, only Estonian is not a Slavic language.

Latvian and Lithuanian are not at all Slavic. There is a branch that contains both Baltic and Slavic languages, but there's also one that contains Albanian and Greek.

Albanian and Greek are both completely separate branches, and both unique on the tree (they don't have common cousins like the others).

There have been some attempts to tie Albanian to Germanic, or Greek, or other branches, but they all have failed.

At some point they all are Indo_european, but they split a way ago.

Re: EuroLLM: LLM made in Europe built to support all 24 official EU languages

#147

Post by pier25 »

deaux wrote: post #99
whimsicalism wrote: post #57

[flagged]

> What's REALLY much more important though if you want to be a part of the AI race and I've posted for years here with @euaccofficial is to make Europe a really extremely attractive place to start and run an AI business. Remove regulatory obstructions and give tax discounts for startups. Let them build a business first that can compete worldwide and once they make enough money (let's say $100M/y), then slowly start a…

> When you talk to most EU business owners, even in tech, the limiting factor isn't regulations.

I have a tech startup in Estonia and I agree. To me the biggest limiting factor is lack of funding.

Re: EuroLLM: LLM made in Europe built to support all 24 official EU languages

#148

Post by lukan »

clickety_clack wrote: post #127
deaux wrote: post #99

Earlier quoted context omitted.

> What's REALLY much more important though if you want to be a part of the AI race and I've posted for years here with @euaccofficial is to make Europe a really extremely attractive place to start and run an AI business. Remove regulatory obstructions and give tax discounts for startups. Let them build a business first that can compete worldwide and once they make enough money (let's say $100M/y), then slowly start a…

It’s probably the people who didn’t start a business in the EU that you want to talk to. Like, I’m European, but I started my company in the US because everything is so much easier here.

What would you want to see changed to consider coming back?

Re: EuroLLM: LLM made in Europe built to support all 24 official EU languages

#149

Post by outside1234 »

rsynnott wrote: post #67
przemub wrote: post #43

Earlier quoted context omitted.

Each EU country nominates one official language for the EU, otherwise we'd have Catalan, Breton, Kashubian and many more.

They could get Austria to do it, as it presumably has a spare slot.

This raises an interesting question. Is there only one dialect of German in the LLM? My understanding is that the German German and Austrian German dialects are significantly different.

Re: EuroLLM: LLM made in Europe built to support all 24 official EU languages

#150

Post by FinnKuhn »

numpad0 wrote: post #132
loandbehold wrote: post #10

Aren't all frontier models already able to use all these languages? Support for specific languages doesn't need to be built in, LLMs support all languages because they are trained on multilingual data.

Not natively, they all sound translated in languages other than English. I occasionally come across French people complaining about LLMs' use of non-idiomatic French, but it's probably not a French problem at all, considering that this effort includes so many Indo-European languages.

I can at least also confirm this for German. Here is one example that is quite annyoing:

Chat GPT for example tends to start emails with "ich hoffe, es geht dir gut!", which means "I hope you are well!". In English (especially American) corporate emails this is a really common way to start an email. In German it is not as "how are you" isn't a common phrase used here.

Post reply on HN