For those curious, the 24 official languages are Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Irish, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Slovak, Slovenian, Spanish, and Swedish.
Maltese, interestingly, is the only Afro-Asiatic derived language.
Hungarian, Finnish, and Estonian are the three Uralic languages.
All the others are Indo-European, Greek being the only Hellenic one, Irish the only Celtic, the rest are Baltic, Slavic, Italic, or Germanic.
(I originally used the term Balto-Slavic, though I was unaware of some of the connotations of that term until just now. Baltic and Slavic do share a common origin, but that was a very very long time ago)
Arabic, even. An outlier, as it is AFAIK the only arabic dialect that is not written with the arabic alphabet. Also it's far removed from other arabic dialects.
As a Brit I feel very at home when hearing/reading Dutch and Frisian. It’s a reminder that England and the Low Countries share a lot of close history all the way back to Anglo-Saxon times; of being fishers, traders, burghers and mercenaries moving around the North Sea chasing opportunities, spreading and augmenting languages.
“Brea, bûter en griene tsiis is goed Ingelsk en goed Frysk”
Not only on the language but also in gastronomy and architecture. When I see old towns in UK I usually think about Dutch towns but just without any biking infrastructure.
> However modern standard Dutch (Nederlands, Hollands) is based upon Franconian, rather than Saxon dialects.
> Some of these [Old Saxon] speakers took part in the Germanic conquest of England in the fifth century AD. While it is not true that English and Plattdeutsch derive completely from the same source, the Old Saxon input into Anglo-Saxon was of primary importance and this linguistic group contributed greatly to the Anglo-Saxon dialects which our English forefathers spoke.
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.
When spoken? Almost certainly. But I think they mostly write in Hochdeutsch, especially in formal contexts, at least that I've seen (private chats/etc are a totally different matter), so I don't foresee any major issues there.
Austrian standard german is slightly different from the German variant, even when written. The differences are pretty minor, though, so it’s very possible to have a relatively long text without being able to tell which one it actually is (especially when potatoes are not referenced in it).
My German teacher always claimed that Swiss German and German German (Hochdeutsch) were so different that she needed subtitles to understand it, and she didn't understand why they weren't considered separate languages.
They really are very, very different. Knowledge of one helps with the other, but it's far more than just "a couple of weeks to adjust to the accent", for example.
EDIT: It's worth noting that this is mostly a spoken thing, AIUI - most formal/semi-formal writing would be in Hochdetusch rather than a local dialect.
Is English a legacy official language then from the time the UK was a member (I‘m guessing Ireland nominated Irish instead of English). Aside it feels very un-EU to push this limitation, as I was under the assumption that EU was all about celebrating (European) diversity.
Including the nasty political side-show that is Ulster Scots - literally only brought in as a chilling effect 'whataboutism' to diminish support when Irish speakers ask for language rights in Northern Ireland.
Well Scots is a real language. As much as English or any other. Whether enough people speak it especially in NI to justify it having an official status and such is another matter.
Haha I just added it as a fun fact, I don't actually believe folks will need to start retraining things, or that this is likely to be at the top of the priorities list for anyone. Party programmes are aspirational anyway.
I recently discovered Maltese existed, and started learning it that day. I find it such an awesome language, and not just because of the letter Ħ
I do wonder what natives think and feel about the longevity of their language? What is taught in schools at what ages (assuming English is in the mix somewhere). Is there enough media in Maltese for Malti to go about the moderns at fully in Maltese? It’s shockingly hard to find any information on Maltese, and even harder to find content.
I’m not sure if’s dying out, or in danger thereof; if there are preservation efforts, or if there is no need.
Native Maltese speaker here. It is thought in schools alongside English, with both being official national languages. Most people locally, that are not foreign born or immigrants speak the language, and it is used in most households as the main language. But everyone grows up bilingual, as English is essential for most everything else that we do as a nation.
Not a question, but - Tatoeba could use your help! It is an open source (both code and data) dataset of parallel sentences and their Maltese data is very lacking. Also it’s pretty fun to just translate a bunch of random sentences into a language you speak. :-)
I don't have much personal experience in attempting to communicate with arabic speakers. From others I have heard Lebanese arabic is the closest and you can have a passable conversation.
Not sure which Tunisians are claiming this but they'd definitely need a lot more than minimum effort. Maltese split off from Arabic around 1k years ago. The two languages sound pretty different, and are written with different alphabets.
How are loan words viewed? Do businesses work in Maltese? Are monolingual speakers of the language regarded differently than those fluent in English? Do young people in Malta listen to Maltese music?
Maltese has been loaded with loan words since forever. 5 points if you can guess where bonġu, bravu and mappa come from. At some point there was some literary council for the language that decided that any new loan words should just be spelled phonetically. Computer became kompjuter.
Businesses do work in Maltese and English. Both are official languages. Its quite rare to encounter a business that deals near exclusively in Maltese. Many prefer Maltese but will fall back to english where necessary.
Regarding monolignual speakers, I think theres a lot of stereotypes for maltese only, english only and code switchers. I think its all a bit silly... So as long as communication can happen I don't fuss.
On Maltese music... There's a lot of low ish quality music then there's a few absolute gems. Look up The Travellers, Lapes, Jon Mallia on YouTube/Spotify.
I can concur. All older words (think any word that was needed since the older generations), are Arabic based. All the numbers, all older verbs etc. 'Newer' words are latin based.
Interesting, but I get the impression that ubiquitous English loan words in seemingly every language is a lot different than loan word patterns of the past. Do you think? Maybe not?
I'm actually really curious about everyday usage of the language; is code switching between English and Maltese more common than Maltese on its own? I've seen a few online communities where the vocabulary switches between Maltese and English very often which is interesting but I wonder how much of that is just online / written versus everyday speech.
Depends on where you live and how you were brought up, but for the most part code switching is default.
There was a point about 7 years ago when the overton window shifted to "speak english to strangers first" because of a large influx of foreigners who did not know the language. Since then I've met foreigners who have better Maltese than some natives.
Older folks & geriatrics will sometimes be surprised when they assume someone is foreign and they turn out to be Maltese. "int Malti??" is a statement I get often because I don't look Mediterranean despite being born here.
How is "Marsaxlokk" really pronounced? I've heard that word a few times, but never from a native. Google translate can't help me here, as it doesn't seem to have Maltese text-to-speech.
From what I have heard, Lebanese Arabic is the closest, and still pretty far. Passable conversation is possible.
Maltese is definitely its own language. Arabic roots are there (theres a Semitic joke in there ) but it isn't arabic anymore. Its written left to right with a variant of the english alphabet.
I know that the reverse understanding isn't too bad from chatting with a Saudi-born member of staff on holiday in Malta.
I don't think anyone would seriously consider it a dialect of Arabic though with its completely different alphabet and half the vocabulary and morphology coming from Italian languages/dialects, even if Malta hadn't spent the best part of a millennium trying very hard not to become part of the Arab world
Seems like the model isn't limited to those though, from the paper:
> as well as some additional relevant languages (Arabic, Catalan, Chinese, Galician, Hindi, Japanese, Korean, Norwegian, Russian, Turkish, and Ukrainian).
The paper also goes into detail on training set sources, which I feel like a curation thereof might be considered the main contribution of this publication?
I think some people get touchy about them being lumped together if their last period of commonality (per the article) was 1400 BCE. For comparison, I believe all the Slavic languages were mutually intelligible around 1200 AD. But much more recently than this, in the last few centuries, there have been notable attempts by east slavs to absorb the Baltic language cultures and deny them.
Depends on your standards, too. Even today, any pair of slavic speakers should have a head start in understanding each other. Put them next to each other for a month and they should be talking, at least about basic everyday things.
Well there is if you go far enough. It's just the question when did they split off from each other. However there is no question that Baltic and Slavic are more closely related to each other than any other non extinct Indo-European languages.
The fact they they are the closest surviving relatives on it own doesn't mean it makes sense to group them together (i.e. Italo-Celtic is also a theorized subgroup in a similar way but nobody is disputing that Celtic and Italic languages evolved into distinct groups).
Then there is a huge amount of missing links and unknown unknowns. e.g. Thracian and Dacian probably were also pretty close to Baltic or Slavic (maybe even closer to Baltic than Slavic is but we don't know enough about them to make any conclusive claims at all... but we at least know these languages existed)
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.
I regret being that loose with the designation :), Romanian and Hungarian are valid counter arguments.
In my mind, I was thinking of the belt of countries between Russia and Central Europe, starting from the Baltics down to the Balkan (excluding Greece).
Even by your definition, I can count at least seven countries where the official language is not Slavic. And that's not even including all the Altaic, Romance and other assortment of regional languages, many of which have some sort of official status.
Some of my fellow Romanians will also claim they're Central European, but in my mind, all the ones I listed are Eastern European countries. I'd even include Turkey and Kazakhstan in there, part of the latter is to the West of the Urals, which is what we normally consider the border between Europe and Asia.
“There’s a question” implies that there is a ground truth that might be discovered to resolve this rather than simply a clash of different purely arbitrary definitions of the same terms.
Ah, yes, how could I forget! As a side note, though also Finno-Ugric then similarity in sound and appearance from Finnish or Estonian at least appears very far.
The Visegrad 4 (Poland, Czechia, Slovakia, Hungary)are generally taken to be "Central European". The strict East/West division is largely a product of the Cold War and the Iron Curtain.
Plenty of wrong here, considering Lithuanian and Latvian are utterly unintelligible to slavs, save for loanwords, but Slavic languages between themselves retain some level of intelligibility, which even spawned two competing constructed languages.
I was thinking about separating the two groups when I was writing this but was afraid of getting too verbose, though in retrospect that probably would have made more sense regardless of the historical lineage. My apologies if this came off as inconsiderate.
I updated my original comment, and learned a good amount about that dispute as a result, so thanks for calling it out.
Norwegian is not on this list, because in fact no country with Norwegian as their national language is part of the European Union at the time of writing.
Flemish is more of a political construct than linguistic, it's a grouping of belgian-dutch the coastal, brabant and limburg language groups with each having their own regional dialects.
It's more than political. In speaking Flemish is to Dutch as UK English is to US English. In writing however there is no difference in spelling, but there is a difference in word choice.
Now, being from Belgium, even within that small part of the country where everybody is supposed to speak Dutch, I genuinely don't understand people from near the coast, which was about 150 miles from where I used to live.
Not sure that should be the qualifier, there might be more people able to speak Basque in the world than Danish, doesn't stop Danish from being well supported.
Quick google points to about 1M Basque speakers in the EU against 5-6M Danish speakers, there's also the fact that Basque is not the only official language in the country it belongs to, and that it's in fact not spoken in the vast majority of the country.
>One of the EU’s founding principles is multilingualism.
>This policy aims to:
>communicating with its citizens in their own languages
>protecting Europe’s rich linguistic diversity
>promoting language learning in Europe
With this in mind, the first intention fails by an enormous margin, given that 95%+ of Spain doesn't speak an iota of Basque, the second is met handily, given the long history of the language, and I'm not sure what to think about the third, any language whatsoever would serve that purpose.
I think those 24 languages reflect all the languages that are official languages at country level.
So for instance, Basque is not an official language of any country (only French in France and Spanish/Castilian in Spain). Belgium's official languages are French, Dutch, and German, "Flemish" is only a local variant of Dutch (Belgian French is also only a local variant of French).
Official is a weird concept though. Turns out Dutch law never really bothered to define an official language, Dutch simply is the de facto standard and is required for a lot of things making it effectively the standard. This makes Dutch Sign Language the only language officially recognised by law. An attempt to recognise Frysian and Dutch as official languages in the constitution failed.
Sweden didn't have an "official" language before the Language Law of 2009. Five minority languages (Finnish, Meänkieli, Romani, Sámi, Yiddish) were officially recognized as such since 1999.
Basque is an official language and declared as such in the Spanish constitution however restricted only to the regions that decide to apply it (Basque Country and Navarra).
If we want to go all legal, I believe that Spanish/Castilian is the only official language of the State, so at country level, with the other "Spanish languages" only official in their respective areas:
Section 3
(1) Castilian is the official Spanish language of the State. All Spaniards have the duty to know it and the right to use it.
(2) The other Spanish languages shall also be official in the respective Autonomous Communities in accordance with their Statutes.
(3) The richness of the different linguistic modalities of Spain is a cultural heritage which shall be specially respected and protected. [1]
In the US, people will resort to fisticuffs, over variants of Spanish. I usually translate into Castilian Spanish, because that seems to be the equivalent of "Vanilla" Spanish. No one is really happy (except the Spaniards), but I'm not accused of favoritism.
English is an official EU language because Regulation 1 Article 1 says so [0] and hasn’t been changed. In practice, English is the most widely used language in EU institutions, so it would be have been silly to remove it after Brexit.
English at this point has stopped culturally belonging to the United Kingdom and whilst one can discus it's not so very moral way of getting there, it's become the bridge language for people of different languages to communicate in, further solidified by the internet.
That said, whenever there is a language selection UI (e.g. at banking machines or institutional websites) in wider Europe that uses flags to represent languages -- probably not a good idea to start with, but very common -- the Irish tricolour should be used to indicate English rather than the UK or USA flags. (although cf Airteagal 8 of Bunreacht na hÉireann).
AFAIK Ireland only listed Gaelic as their official language with UK having English. That caused a bit of a problem during Brexit since technically English wasn't officially an EU language anymore. I guess they resolved it somehow.
Actually nuts to me the degree to which European policymakers do not even begin to understand how to kickstart technologically-intensive industry. Anyone who has seen close-up the results of a "pick the winners" grant-style approach to innovation knows what will go wrong here.
While grant process in EU isn't fun, I think Levels has bit of an ego issues. He mentioned that if he had issues like that on eg X, he would see Elon himself in the replies.
While he is great at converting his influencer status to income in his micro-SaaS projects, I don't think running ad-fueled browser games on state-sponsored super computer should be really aim of these grant programs.
He is 100% right on this one. From personal experience trying to figure out EU. Lawyer bureaucrats manage funds behind red tape clearly meant to be for their pals.
All these while the EU is running out of funds and in a process of de-industrialization. There should be an independent corruption investigation on Brussels.
I took part in application for EU grants a few times and our company group did it many times over the years.
It's bureaucracy, often bordering with stupidity. You may need advisors to navigate all their forms & processes. But it certainly isn't "pals-only" type of deal.
On the other hand - is it harder than getting VC funding? For seasoned founder with reputation - probably. For fresh startup - probably not.
Reviewer (Scientific Expert) for the EU (since 2009) here.
The probability of getting a Horizon Europe grant allegedly (not official stats) is about 8.5% according to some friends, which may seem low. You need to write 70 pages following a Word template and the key goal is to cover answers to a large number of questions. Each proposal gets various grades across a range of dimensions, which get added up and if you obtain at least 13 out of a possible 15 points, you are eligible to get funded, read: "You will get funded if there is enough money." Often, there are several proposals that justly achieve 15/15, and because of that, many prosals that have 14 points and all proposals that have less may not get funded, simply because there just is not enough total funding available to fund all the technically eligible proposals. Having judged many proposals in AI / ML / search / "big data" / language technology etc. I recommend optimizing recall, i.e. aspiring completeness.
The application process is not easy, but you can get help: there are support agency in each member country, free online Webinars to help, hotline help desks as well as an ecosystem of paid consultants that typically charge about 3k€ to vet a proposal for you if you need that kind of service (I never used it).
The process is neutral and conducted professionally and with external oversight (consultants are hired as "rapporteurs" that report on process/procedural integrity in additional to the actual reviewers). I value the research officers of the EC as people of high competency, integrity and motivation (research money is tax payers money so it should be spent carefully).
In comparison, VC (and even more so business angel) funding is achievable with much less formal apparatus, often a short business plan and a convincing slide deck and demo can get people to a partner meeting if the time is right. But the criteria and process are much different, and ideas ready for public research grants are typically too early for VCs (but the EC wants to foster the creation VC-funded startups resulting from the disseminated research).
Right but if we do this with public funds then the narrative shifts to "OMG the EU is so corrupt and stupid, looking they're pouring taxpayer dollars into unproven stuff! They're deindustrializing!!"
The point being that, as soon as public dollars are on the table, people expect perfection. Anything less is waste, fraud, and abuse.
There's literally no winning. Want to make sure the money is allocated right? Bureaucracy. Want to not do that? Waste, fraud, and abuse.
Except that Apple, Intel, Tesla, etc have all received US government investment [1]. TSMC is a product of the Taiwanese state! Government investment can be done well, and seeds excellent companies.
Of course there is red tape. EU funding comes from taxpayer money and we want it to be spent wisely. The red tape is precisely to prevent it from being funneled to pals. EU has funded quite a few free software projects so it's not like the red tape is an insurmountable burden: https://www.ri.se/en/news/blog/europes-digital-future-spells...
I'd also say that their grants aren't unusually burdensome and grantmaking is arms length compared with a lot of other bodies.
Yes, some of the questions are weird, but I'd really rather write a bit confirming that the AI system being developed isn't going to be racist or Skynet than jump through some other hoops that exist (and that absolutely includes VC due diligence). The actual biggest issue with European funds is they get way more competent applications than they can fund anyway.
I'm actually no fan of his, so that's fine. That said, I went to the actual website he was talking about (I'm also an EU citizen) and in this case it is exactly as described and bordering comical.
> 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 adding regulation.
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.
Ironically, China has in some ways a bigger regulatory burden when it comes to software, as there if the government doesn't approve the business is dead in the water. I doubt that Klarna would've gotten off the ground there, for one, I could see them being shut down much earlier there. In the EU only now very slowly are some governments even starting to talk about some weak measures around their business model. But I've never, not once in my life, heard "Chinese software companies can't get off the ground due to the regulatory burden".
The same people who clamor about the EU regulations are the ones who hate on the EU for their protectionist measures against US tech. Yet another bout of irony here - China's software industry has flourished exactly thanks to 10 times stronger protectionist measures against US tech. So has Korea's, and their protectionism has never even been anywhere on the China level, more inbetween EU and China. No, if there's anything that would help, it's much more tech protectionism in the EU.
Pieter Levels is at the end of the day an influencer, not a serious founder.
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.
Where in Europe and where in the US? You probably started one in the easiest US state to do so. Did you try starting one in the easiest EU state? Otherwise we already can't take things very seriously.
Secondly, what's easier besides VC funding? If it's VC funding, the disparity there has nothing to do with regulations - guess how much VC funding the non-EU rest of the world gets.
> but I started my company in the US because everything is so much easier here
Which part is easier? That you have 50 different states with slightly varying laws to consider (e.g. Californian Data protection)? That you have a byzantine system of "benefits" to choose and manage?
And compared to where? Germany or Estonia or Sweden or Spain? The complexities will vary wildly depending on the country (kind of like in the US, where lots of companies pick the state to base themselves in based on the combination of favourable laws and precedents and taxes).
we might be somewhat trending in this direction, but the reality is largely that the US states are pretty identical and have very similar laws on the books. the federal government is in charge of commerce usually.
most laws like CCPA also have some threshold where you already need to be pretty successful for it to apply to you.
for some select industries (biometrics & healthcare), yes you have a patchwork of laws.
Opening a company in Estonia is very cheap but in Spain the manager/CEO needs to be an "autónomo" (like a self-employed tax status). This costs thousands of Euros per year. Something like 2,400-30,000 Euros per year, every year, forever.
> 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 competitive wages. There's a real difficulty in allocating capital to talented humans.
And, at least in Southern Europe, the income tax schedule is so aggressive it's hard to justify continuing working in many of these countries if you are highly talented.
Like, if you can tell me what the induced operator norm from l_2 -> l_2 is - probably you should come to the US and work at a biglab and make bank. What can you do in Portugal, Italy, Spain, etc.??
> Pieter Levels is at the end of the day an influencer, not a serious founder.
Sure, agreed.
I think it is a complete misreading to point to protectionism as the reason for Chinese success, but having a big unified domestic market for consumers along with massive saving rates and capital controls probably does help.
One fairly large factor is that even though English is much more common today, you just can't operate (depending on the product of course) in many countries without having customer support, documentation etc in the local language.
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.
Are you saying that the other 199 non-US countries in the world all have a business unfriendly environment, since every one of them besides China has practically the same amount of software VC funding compared to the US?
All of these purported EU-specific reasons completely ignore that things are the same elsewhere. It's the US that is the outlier.
> I think it is a complete misreading to point to protectionism as the reason for Chinese success, but having a big unified domestic market for consumers along with massive saving rates and capital controls probably does help.
Capital controls are protectionist measures, but anyway, no.
> Okay, what is the limiting factor?
Let's look at which countries have a significant local software industry compared to population size.
- China
- US
- Korea
- You can argue for Japan and India but that's already starting to stretch.
- Yup, effectively no where else. Even in an "out of the way" place like Myanmar everyone uses Meta, with a nice little genocide to show for it. Sure, in Vietnam they use Zalo, and other places have a few other local players. But most of the famous US tech apps are dominant.
Is the EU the outlier here? No. Everywhere else US tech dominates. Meta, Netflix, Apple, Google, Uber, Spotify, Microsoft, Match Group, Paypal, Amazon, and on and on. They don't just dominate the EU, they dominate the world.
Except for the countries I named above, where at least some of the markets that US big tech competes in, instead have bigger local players. And even there, guess what?
Their market share is almost 1:1 linearly correlated to the degree of protectionism in those countries, all the way from China, then Korea, then India/Japan, and then everywhere else! Who woulda thought!
Why does Korea have much less US tech dominance than, say, Germany? Despite German companies theoretically having a big advantage: the German public is 100x more privacy conscious than the Korean one, and much less trusting of US companies.
I can tell you that it's not less regulations; Korea's GDPR is much more onerous than the EU's and so are investment regulations. On every single regulatory aspect, German software startups have it easier. But they were never protected. US tech was allowed to waltz in, dump their products - that's what they did, it's hilarious how now China "dumping" EVs and solar is suddenly an issue when it's exactly the strategy that US tech continues to this day; the AI companies are doing it right now! And the Korean companies were protected. Both by the rules burden, that local companies had to deal with too, along with intentional protectionism.
When it comes to solar and EVs, we all understand that a foreign country dumping their goods kills local industry. It's the exact same with software.
But then half of HN has millions on the bank exactly thanks to the above - this is where all those fat SV salaries have come from - so I do get the lack of desire to understand it.
> Their market share is almost 1:1 linearly correlated to the degree of protectionism in those countries
Seems like you actually believe this. I think our starting points on reality are different enough that we are not going to have a productive conversation, I wish you and other Europeans the best of luck in your protectionism-led growth strategy. Make sure to not discuss it with any pesky macroeconomists who might lead you astray. take care
A big part is that the EU is a collection of countries that (with very few exceptions) have different languages and laws. For a company to serve Spain and France, for instance, it would need to translate everything, hire local lawyers and customer support agents. Considering the much smaller size of the countries (biggest one is 70 million vs 330 million in the US), the opportunity for "unlimited" growth is limited.
This also rebounds in the fact that when an American company makes it big, they have the resources to flood other EU markets and be cheaper/better than the local competition due to economies of scale and money based on their big successful US market. A French company making it big is still small compared to a US equivalent.
Then, there's the capital markets, no denying that. The money being thrown around the US is like nowhere else on the planet. Some of it definitely a bubble / unrealistic, but that doesn't matter. But in part it's because of the size of the total potential market that this is justified.
Education / national mythology also plays a part, I think (this is pure conjecture now). In the US, the "American Dream", "everyone can make it" etc is heavily ingrained. It propagates through the world with the help of Hollywood and other American cultural exports. In most EU countries, there isn't such a heavy emphasis on independence and "pulling yourself up by your bootstraps". "Hustle culture" isn't a thing. So for most people, it isn't something that comes naturally to them to start a company and work 100 hour weeks to be big and rich and successful and famous.
That's not to say there aren't such people, I went to 42 and have been to Station F and know some people in that universe. A decent proportion of my classmates wanted to make their startup and make it big, and some did end up starting their own companies.
> This also rebounds in the fact that when an American company makes it big, they have the resources to flood other EU markets and be cheaper/better than the local competition due to economies of scale and money based on their big successful US market. A French company making it big is still small compared to a US equivalent.
Ding ding ding! When China does it with solar and EVs we call it "dumping". When Uber, OpenAI and Anthropic do it, that term is never ever used. VC funded US techs dumps harder than any Chinese industry ever has.
Tbh, a lot of EU protectionism vs. US tech seems not to keep the competition out. In fact, with the amount of free press US startups get and the size of their coffers, they can simply roll over the local competition in EU markets most of the time.
What it's terribly good at is adding burdens that the US giants don't face early on, slowing down the early growth between 28 fragmented markets. I don't know specifically about how China works, but the question is proving product-market fit, and for that, you need a lot of users fast.
In the EU, it's a different battle country to country as the media environment, the markets, the regulation etc. are all fractured.
Is the point of these policies to pick winners? Or to upskill the creators and stimulate the economy by giving possible entrepreneurs experience Europeans can't get in big tech?
In the US, some ex-Googler might found a startup. Europe doesn't have the equivalent of FAANG. (Europe-wide companies are not quite as easy as US-wide)
Even if the super computer itself "fails", is the goal actually the secondary impacts to the economy?
(And in the US, we do our own fair share of picking winners / losers, especially in the current regime)
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.
I don't think what you're saying is inconsistent with what I'm saying. I think you are making a big deal out of the difference between state investment funds and subsidized GPUs but I think they basically work by similar mechanisms.
> 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.
Does government offering matched funding to investors actually help startups who are struggling to find (any) funding? If a startup can't find (any) funding, matching is irrelevant.
> 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.
Umm. I'm not really convinced that the political elites in Europe understand how to do any of this stuff well.
I literally quoted the paragraph from this link in the tweet I provided: Edit: lol, I didn't, I quoted it from a policy document, not from press release. However, my point stands:
--- start quote ---
Apply AI Strategy
The Apply AI Strategy aims to harness AI's transformative potential by driving adoption of AI across strategic and public sectors including healthcare, pharmaceuticals, energy, mobility, manufacturing, construction, agri-food, defence, communications and culture. It will also support small and medium-sized enterprises (SMEs) with their specific needs and help Industries integrate AI into their operations.
--- end quote ---
I also quoted a paragraph from a document I will find when I'm not on mobile.
> Seems like your quote was very misleading to me, so no your point doesn't stand
My quote: Cluster: for public benefit, cutting edge research in biotech, medical, robotics.
Literal quote from your link: The Apply AI Strategy aims to harness AI's transformative potential by driving adoption of AI across strategic and public sectors including healthcare, pharmaceuticals, energy, mobility, manufacturing, construction, agri-food, defence, communications and culture.
You: your quote was misleading.
I'm sorry, I don't have the time or the patience with willfully ignorant and blind people getting their interpretations from AI slop engagement farmers.
By funding AI slop, you're funding AI slop, not AI research, or, quote, "drive adoption of AI across strategic and public sectors including healthcare, pharmaceuticals, energy, mobility, manufacturing, construction, agri-food, defence, communications and culture"
EU grifting is so much worse than even the most brazen Trumpian crypto pump n' dump.
Geniunely repugnant. Atleast the Trump admin has the decency to pump everyones 401k...
I'm trying to figure out why it bothers me so much. I think its because the EU are such unbelievable losers in everything they do. they can't even grift, thats how useless they are. they can't even steal properly. its so undignified, and offensive to the senses.
Wouldn't go that far. EU policymakers have good intentions, I believe - but ultimately are products of their environment and cultural inclination.
The EU is such a bizarre place because they treat capital and entrepreneurs with such massive distrust, but never really bothered getting rid of the quasi-static entrenched hierarchies from feudalism? Like I'll go to the UK or France and there will just be massive swathes of land owned by the nobility or 'former' nobility? Maybe start there but let your high-value human capital earn a good wage?
You are wrong in that you think the hierarchies stem specifically from feudalism, but you are absolutely correct in that these hierarchies exist and are deeply entrenched. Sweden and Germany have one of the lowest percentages of self-made vs. inherited fortunes in the western world. Actually some tax policies in the US enable much more upward mobility, such as real estate taxation and 401k-like vehicles.
>The EuroLLM Team brings together some of the brightest minds in AI including Unbabel, Instituto Tecnico Lisbon, the University of Edinburgh, Instituto de Telecommunicacoes, Université Paris-Saclay, Aveni, Sorbonne University, Naver Labs, and the University of Amsterdam.
>Europe is the only continent in the world to have a large public network of supercomputers that are managed by the EuroHPC Joint Undertaking (EuroHPC JU). As soon as we received the EuroHPC JU access to the supercomputer, we were ready to roll up our sleeves and get to work. We developed the small model right away and in less than 6 months the second model was ready.
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.
No, that's not how training works. It's not just about having an example in a given language, but also how many examples and the ratio of examples compared to other languages. English hugely eclipses any other language on most US models and that's why performance on other languages is subpar compared to performance on english.
Ratio/quantity is important, but quality is even more so.
In recent LLMs, filtered internet text is at the low end of the quality spectrum. The higher end is curated scientific papers, synthetic and rephrased text, RLHF conversations, reasoning CoTs, etc. English/Chinese/Python/JavaScript dominate here.
The issue is that when there's a difference in training data quality between languages, LLMs likely associate that difference with the languages if not explicitly compensated for.
IMO it would be far more impactful to generate and publish high-quality data for minority languages for current model trainers, than to train new models that are simply enriched with a higher percentage of low-quality internet scrapings for the languages.
I have never noticed any major difference in performance of ChatGPT between English and Spanish. The truth is that as long as the amount of training data of a given language is above some threshold, knowledge transfers between languages.
I keep hearing that LLMs are trained on "Internet crap" but is it true? For instance we know from Anthropic copyright case that they scanned millions of books to make a training set. They certainly use Internet content for training but I'm sure it's curated to a large degree. They don't just scrap random pages and feed into LLM.
If it's publicly available data, books and research, I can assure you the big models have already all been trained on it.
European culture is already embedded in all the models, unless the people involved in this project have some hidden trove of private data that they're training on which diverges drastically from things Europeans have published publicly (I'm 99.9% positive they don't...especially given Europe's alarmist attitude around anything related to data).
I think people don't understand a huge percentage of the employees at OpenAI, Anthropic, etc. are non-US born.
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.
Meh, it depends a lot on the dataset, which are heavily skewed towards the main languages. For example they almost always confuse Czech and Slovak and often swap one for the other in middle of chats
You can also correctly bias your sampling so that when selecting new training instances each language is chosen equally. Generally the diversity of data is good, unless that data is "wrong" which, ironically, is probably most of the internet, but I digress.
The difference ia larger than let's say just a "dialect". They really are different languages, even though we generally understand each other quite well (younger generations less so). I've heard it's about as different as e. g. Danish and Swedish - not sure if that comparison is helpful.
From what I am aware of, LLM capability degrades once you move out of English, and many nation states are either building, or considering the option of building their own LLMs.
I was thinking the same, why are so many superior models coming from only countries like US and China. And why are European countries not in the list other than France with Mistral. Why are so few companies in India, Japan, South Korea even close to a promising new model like what Chinese companies did ?
"Why" is a fair question but are you surprised? Europe is consistently behind in tech.
Europe has about 1.3 times the population of the USA and about 75% of the GDP yet EU tech output is a very small percentage of US tech output. We are not talking about 70, 50, 30, or even 20%. It's a drop in the bucket.
>The seven largest U.S. tech companies, Alphabet (Google), Amazon, Apple, Meta, Microsoft, Nvidia, and Tesla, are 20 times bigger than Europe’s seven largest, and generate 10 times more revenue.
"Why" is a good question, but I definitely wouldnt expect significant competition in LLMs from Europe based on the giant tech disparity. Having 1 non-cutting edge model that isn't really competitive is pretty much what I would expect.
> The seven largest U.S. tech companies (...) are 20 times bigger than Europe’s seven largest, and generate 10 times more revenue.
I'm going to guess that this part is intentional. Europe tends to be more aggressive in enforcing antitrust laws. Economically, Europe's goal isn't to have the biggest companies but to have more smaller companies.
So you're not going to get companies like Google, but you will get companies like Proton, Spotify, Tuta, Hetzner, Mistral, Threema, Filen, Babbel, Nextcloud, CryptPad, DeepL, Vivaldi, and so on.
>I'm going to guess that this part is intentional. Europe tends to be more aggressive in enforcing antitrust laws. Economically, Europe's goal isn't to have the biggest companies but to have more smaller companies.
So is your hypothesis that the total market cap of EU tech companies is something like 50,60,70, etc. % of total US tech marketcap? Something significantly different than the ~10% implied by that figure (largest us companies 10x largest EU companies). And it's just more broadly distributed?
European companies are smaller on average and less likely to go public in general, so market cap comparisons don’t show the whole picture. Growing big is less often seen as a goal than in the US. “Megacaps” aren’t necessarily considered a healthy thing to have.
I don't see any sense in which the EU has fewer capabilities. It has, say, a smaller number of businesses with smaller market dominance.
It isnt clear to me what capability the EU would gain by having a monopolist social network, a monopolist search engine, a monopolist advertising trader
Also, commercial software is consistently behind from open source.
I only use open source LLMs for writing (Qwen 32b from Groq) and open source editor of course, Emacs.
If some people can write better using commercial LLMs (and commercial editors), by all means, but they put themselves at a disadvantage.
Next step for me, is to use something open source for translation, I use Claude for the moment, and open source for programming, I use GPT curently. In less than a year I will find a satisfying solution to both of these problems. I haven't looked deep enough.
As a European citizen I think it boils down to access to the capital. EU/EEA is not a country and the market is sort of fragmented. The big players are UK, France, Germany, everyone else does not have the same access to money as say in the US. Folks want to do it but there is a glass ceiling. Hence you have these collabs among large institutions to tap into funds such as from Horizon which are academic in nature and do not translate well into products.
You can easily fit below 10 billion for the whole datacenter, then you only pay for electricity + maintenance + staff. 100k GPUs cost a few billion USD, that's more than enough to train frontier models, run experiments, and serve models in the EU to start. Look at what xAI did and how much it cost them and it's more expensive to do in US than in EU.
EU made a >900 page law about AI and patted themselves on the back for being "the first to regulate AI" (which was not even true, China had an AI law before and it's two pages long).
This cannot be stressed enough. In my experience working in multiple tech startups in Germany, the power compliance, legal and all other 2nd line has over engineering is quite immense. Most of the time they act as a hindrance for innovation rather than a supporting factor.
This AI law is a clear example of that. Pencil pushers creating more obstacles for the sake of creating more obstacles rather than actually taking a pragmatic approach.
It's strange, my real life experience is very different than yours. Unless you're training AI to do something shady, it's really no bother at all. In fact, most of what the AI Act requires, you have to do anyway for a good model card.
Because the value of these models is (actually) yet to be proven. Why saturate the market with something that we already have at least one of and others are selling as a service? No model provider (including the "big ones" like OpenAI) has been able to produce a viable business case. They're all literally running on government deals and investor money.
Does it even make sense? Just use the American or Chinese ones, adjust
As needed. Where’s the point in spending millions to build
The same thing or worse
I'm somewhat skeptical of taxpayer funded innovation. Seen a few Horizon grants from the side, as a citizen I'd prefer to not pay for them, but unfortunately can't opt out.
The vast majority of US discoveries are by immigrants using taxpayer money. AKA scientists at universities. Your media likes to give credit to the companies, but generally the companies only apply things, they rarely create new science these days.
My experience with government funding is that they apply something and won't even try to sell it because selling is hard: you don't want to know that the thing you built is lacking nor that the competition is better. Especially the academic types don't. Yet I'm paying for these guys. Also, by funding the academics they won't even need to go to the job market.. But as I paid for their education I thought I was buying people who create value.
Perhaps the above is rather harsh and it's "not that bad", my subjective experience nevertheless.
Much of the neural network work was funded by Canadian Universities, and commercialized by US companies. Even if you look just at the "Attention is all you need" paper, which is primarily by authors working at Google, most of those authors come from academia and are immigrants.
Vaswani is an Indian born computer scientist, Shazeer is US, Parmar was born in India, Uszkoreit was born in Germany, Jones was born in the UK, Gomez is British-Canadian, Kaiser is a Polish computer scientist, and Polosukhin is Ukrainian.
Almost all of these people have PhDs and Master degrees. The ROI on academia is vast for society, including European universities. The thing the US does well is capitalize on that education, and sadly also try to steal credit for it as "American exceptionalism." If Europe and other countries learn how to keep their academics and get them working in local industries, America's edge will evaporate overnight.
A major factor in European academics moving to the US is that top US institutes can charge a small fortune, and some of that gets reflected in academic salaries. Interesting move by the US government to try to put them off...
The wider availability of capital is a bigger deal though. "Attention is all you need" is available to people on other continents to read, but a computer scientist in Europe that understood exactly how big transformers were going to be and why had less chance of funding than a webdev in California with a pitchdeck full of cliches and me-too GPT wrapper for an industry they'd barely touched does today.
> I'm somewhat skeptical of taxpayer funded innovation... as a citizen I'd prefer to not pay for them, but unfortunately can't opt out.
There are a few variables here but at this point in time, private-funded innovation isn't different by much and all things considered, the difference isn't in its favor.
I wouldn't mind actually/visibly productive companies taking these grants. But I've also seen mostly research-focused (nominally) private companies who mostly live off of science grants, who don't produce nor sell much - because they don't have to.
Good to distinguish between publicly funded research models (like this one) and commercial ones (like Mistral in France). What are the chinese and usa public research models like?
I propose an European AI-only "NASA" style agency that would have a frontier LLM-"Apollo Program" goal. It would subcontract the several blocks it needs across EU member states.
Would you prefer European AI sovereignty with 15% overhead costs from geographic distribution, or 100% dependence on Nvidia/OpenAI with zero European industrial base?
That’s a cool idea — training a multilingual model like that is ambitious. But I’m curious how well it’ll actually handle smaller EU languages compared to English or French. If it truly nails those, that’s a big win for accessibility.
I asked a Finnish person how good an answer about the language example from ChatGPT was. It turned out to be a hallucination, a confidently sounding nonsense.
The quality of internet trained models degrade very fast with language material size
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 <language>: "
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
I used the 9B Instruct version, from the small models, it was the one with the best Latvian knowledge out there, bar none. GPT-OSS 20B and Qwen3 30B A3B and similar ones weren't even close.
That said, the model itself was a little bit dumb and not something you'd really use for programming/autocomplete or tool calling or anything like that, which also presented some problems - even for processing text, if you need RAG or tool server calls, you need to use something like Qwen3 for the actual logic and then pass the contents to EuroLLM for translation/formatting with the instructions, at which point your n8n workflow looks a bit messy and also you have to run those two models instead of only one.
Meanwhile, the best cloud model for Latvian that I've found so far was Google Gemini 2.5 Pro, but obviously can't use cloud models in certain on-prem use cases.
It seems there is some weird grouping of the language data which LLM cannot distinguish well. I wonder if it is the same for other similar languages like scandinavian or western slavic
It seems like it, in most ways, it would be bad to train on 24 separate languages. That's just 24 partitions to the data. Seems really inefficient and better to simply train in the biggest (english) and translate.
I do think this will introduce some biases that correlate with the English language. It would be interesting to see more specifically what this means. But regardless, I don't think you can produce a competitive model with such a large subdivision of training data.
If you train a model on multiple languages, you can use the model itself for translation. As well as allowing the model to naturally respond in the user's language.
nah, it's better to train on all languages. 24 partitions? you are gravely underestimating these models and how they represent things in their latents... transfers easily
EU officials should create an environment where abundant private companies can afford to put out many great open models instead of funding some selected individuals with taxpayer money.
What good does it do by having only include formal languages?
For example there’s no Russian, while there’s now at least 8 million ethnic Russians living in Europe.
As the aim of EuroLLM is to provide EU citizens with powerful and useful AI tools, it is critical that the model can also translate and answer questions in other European and non-European languages. With this in mind, we added support for 11 additional languages (Arabic, Catalan, Chinese, Galician, Hindi, Japanese, Korean, Norwegian, Russian, Turkish, and Ukrainian).
tbf, the USSR fought the Nazis mainly because they didn't have much choice after Nazis turned on them a little while after they'd teamed up with those ideological enemies to invade Poland, so it's not like they hadn't put the effort into being on the wrong side of history :)
Surely that would need to be both input and output. But even then you could easily get lost in translation as the intent in one language might mean slightly different thing in another. Thus you could get subpar results.
They almost exclusively compare their model to prior models from 2024 or older and brag about "results comparable to Gemma-2-9B". I'm not sure what I expected. The eurollm.io homepage states "EuroLLM outperforms similar-sized models", which just seems like a lie for all practical purposes
An overly charitable interpretation is that EuroLLM isn't a reasoning model and has minimal post-training, so they sought out comparisons to such models (they're still ignoring reasoning models that have non-reasoning modes)
> They almost exclusively compare their model to prior models from 2024
As another comment here noted, the title is missing (2024) - this model was released almost a year ago, last December, so it's not surprising that that's the models they compare to.
It's more like the default is to be ranked near the bottom unless your comment gets traction during the brief window of time it is ranked first for being new. Seeing your comments go splat after that window expires is not some nefarious conspiracy..
I thought research grants were to make novel discoveries, not to replicate what industry has long done. Unless we are at the point where we study US as an alien civilization
I prefer proprietary LLMs that are actually good products - byproducts of free market competition (capitalism), instead of products created from govt initiatives that lead nowhere (good).
As expected, Europe finally catches up to 2024 and launches an LLM that barely competes against the heavyweights.
The US and China are running rings around Europe.
Mistral is an exception as it was funded by US VCs and they are a great example showing that without VC funding, Mistral would have been begging to the EU for a microsopic grant to train a LLM worse than Llama.
Mistral is pretty much toast? Their models perform poorly and I'm not sure why anyone would use them. Maybe there is a catching up point somewhere in the future, hopefully.
Every country needs a few plumbers and carpenters whether or not they are at the forefront of technology. Some money must be spent to give academics work to do so they can sharpen up their skills and perhaps teach the next set of students who might be more commercial
If you could find one for the money, if they were happy to teach in the long term. If it wasn't better to have N for the price of 1. In other situations of import substitution I'm pretty sure people try to develop their local talent in addition to buying in experts.
Of course catalan isn't in the list. 10 million speakers that don't matter to the European Union. EU likes our productivity but squanders our rights. We are 2nd class citizens.
Now let's wait for the people saying "Spain" could change this. Hypocrites.
For those curious, the 24 official languages are Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Irish, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Slovak, Slovenian, Spanish, and Swedish.
Maltese, interestingly, is the only Afro-Asiatic derived language.
Hungarian, Finnish, and Estonian are the three Uralic languages.
All the others are Indo-European, Greek being the only Hellenic one, Irish the only Celtic, the rest are Baltic, Slavic, Italic, or Germanic.
(I originally used the term Balto-Slavic, though I was unaware of some of the connotations of that term until just now. Baltic and Slavic do share a common origin, but that was a very very long time ago)
> Maltese, interestingly, is the only Afro-Asiatic derived language.
It's Semitic, to be precise.
https://en.wikipedia.org/wiki/Semitic_languages
Arabic, even. An outlier, as it is AFAIK the only arabic dialect that is not written with the arabic alphabet. Also it's far removed from other arabic dialects.
Tomorrow there are elections in the Netherlands, and two parties are proposing adding Frysian to that list: https://neerlandistiek.nl/2025/10/kies-voor-taal/
Best get to retraining those models.
AFAIK, they are trying to get Frisian added to the "European Charter for Regional or Minority Languages", not the official language list.
They get certain recognition, but they are not official in Europe. For example, just from Spain there are 13 languages on that list.
As a Brit I feel very at home when hearing/reading Dutch and Frisian. It’s a reminder that England and the Low Countries share a lot of close history all the way back to Anglo-Saxon times; of being fishers, traders, burghers and mercenaries moving around the North Sea chasing opportunities, spreading and augmenting languages.
“Brea, bûter en griene tsiis is goed Ingelsk en goed Frysk”
Not only on the language but also in gastronomy and architecture. When I see old towns in UK I usually think about Dutch towns but just without any biking infrastructure.
> However modern standard Dutch (Nederlands, Hollands) is based upon Franconian, rather than Saxon dialects.
> Some of these [Old Saxon] speakers took part in the Germanic conquest of England in the fifth century AD. While it is not true that English and Plattdeutsch derive completely from the same source, the Old Saxon input into Anglo-Saxon was of primary importance and this linguistic group contributed greatly to the Anglo-Saxon dialects which our English forefathers spoke.
[1]: http://www.plattmaster.de/plattoew.htm
If you've ever read anything written in old English, it's a even closer to Dutch.
Before the Dutch arrived would it have been something like Welsh that was spoken in England?
Each EU country nominates one official language for the EU, otherwise we'd have Catalan, Breton, Kashubian and many more.
Well, this was 4 days ago, Spain in talks with Germany regarding the addition of official languages:
https://www.politico.eu/article/catalan-basque-galician-boos...
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.
When spoken? Almost certainly. But I think they mostly write in Hochdeutsch, especially in formal contexts, at least that I've seen (private chats/etc are a totally different matter), so I don't foresee any major issues there.
Austrian standard german is slightly different from the German variant, even when written. The differences are pretty minor, though, so it’s very possible to have a relatively long text without being able to tell which one it actually is (especially when potatoes are not referenced in it).
My German teacher always claimed that Swiss German and German German (Hochdeutsch) were so different that she needed subtitles to understand it, and she didn't understand why they weren't considered separate languages.
They really are very, very different. Knowledge of one helps with the other, but it's far more than just "a couple of weeks to adjust to the accent", for example.
EDIT: It's worth noting that this is mostly a spoken thing, AIUI - most formal/semi-formal writing would be in Hochdetusch rather than a local dialect.
If Switzerland was in the EU, it would certainly be made a separate official language.
They are in fact considered separate languages.
Try dutch, it is combination of German and English!
Is English a legacy official language then from the time the UK was a member (I‘m guessing Ireland nominated Irish instead of English). Aside it feels very un-EU to push this limitation, as I was under the assumption that EU was all about celebrating (European) diversity.
Still an official language, thankfully. Officially, because of Cyprus.
Malta and ireland
Including the nasty political side-show that is Ulster Scots - literally only brought in as a chilling effect 'whataboutism' to diminish support when Irish speakers ask for language rights in Northern Ireland.
https://www.reddit.com/r/northernireland/comments/1fivtob/no...
Well Scots is a real language. As much as English or any other. Whether enough people speak it especially in NI to justify it having an official status and such is another matter.
This completely ignores the history of published writing in Ulster Scots going back centuries.
Should be noted- the Netherlands can't unilaterally make changes. Spain has been trying to push for languages to be added and hasn't had luck.
Haha I just added it as a fun fact, I don't actually believe folks will need to start retraining things, or that this is likely to be at the top of the priorities list for anyone. Party programmes are aspirational anyway.
Just do a 50:50 mix of the German and Dutch model weights.
Oops, accidentally made the model speak Limburgish.
I read, write and speak Maltese, AMA if you are curious about the language.
I recently discovered Maltese existed, and started learning it that day. I find it such an awesome language, and not just because of the letter Ħ
I do wonder what natives think and feel about the longevity of their language? What is taught in schools at what ages (assuming English is in the mix somewhere). Is there enough media in Maltese for Malti to go about the moderns at fully in Maltese? It’s shockingly hard to find any information on Maltese, and even harder to find content.
I’m not sure if’s dying out, or in danger thereof; if there are preservation efforts, or if there is no need.
Native Maltese speaker here. It is thought in schools alongside English, with both being official national languages. Most people locally, that are not foreign born or immigrants speak the language, and it is used in most households as the main language. But everyone grows up bilingual, as English is essential for most everything else that we do as a nation.
Not a question, but - Tatoeba could use your help! It is an open source (both code and data) dataset of parallel sentences and their Maltese data is very lacking. Also it’s pretty fun to just translate a bunch of random sentences into a language you speak. :-)
https://tatoeba.org/
Tunisians claim they can understand Maltese with minimum effort, is it reciprocal? How close is Maltese to arabic / tunisian dialect ?
I don't have much personal experience in attempting to communicate with arabic speakers. From others I have heard Lebanese arabic is the closest and you can have a passable conversation.
Not sure which Tunisians are claiming this but they'd definitely need a lot more than minimum effort. Maltese split off from Arabic around 1k years ago. The two languages sound pretty different, and are written with different alphabets.
Also lots of influence from Italian and English.
How are loan words viewed? Do businesses work in Maltese? Are monolingual speakers of the language regarded differently than those fluent in English? Do young people in Malta listen to Maltese music?
Maltese has been loaded with loan words since forever. 5 points if you can guess where bonġu, bravu and mappa come from. At some point there was some literary council for the language that decided that any new loan words should just be spelled phonetically. Computer became kompjuter.
Businesses do work in Maltese and English. Both are official languages. Its quite rare to encounter a business that deals near exclusively in Maltese. Many prefer Maltese but will fall back to english where necessary.
Regarding monolignual speakers, I think theres a lot of stereotypes for maltese only, english only and code switchers. I think its all a bit silly... So as long as communication can happen I don't fuss.
On Maltese music... There's a lot of low ish quality music then there's a few absolute gems. Look up The Travellers, Lapes, Jon Mallia on YouTube/Spotify.
I can concur. All older words (think any word that was needed since the older generations), are Arabic based. All the numbers, all older verbs etc. 'Newer' words are latin based.
Interesting, but I get the impression that ubiquitous English loan words in seemingly every language is a lot different than loan word patterns of the past. Do you think? Maybe not?
I don't have much of an opinion I suppose english language cultural dominance has meant that newer words are just imported rather than adapted
Yes, there's plenty of Maltese spoken and listened to.
I was surprised to hear Maltese radio stations played in taxis, while visiting Malta just a few weeks back
The point of my question was to ask someone who lives there, not someone who visited
I'm actually really curious about everyday usage of the language; is code switching between English and Maltese more common than Maltese on its own? I've seen a few online communities where the vocabulary switches between Maltese and English very often which is interesting but I wonder how much of that is just online / written versus everyday speech.
Depends on where you live and how you were brought up, but for the most part code switching is default.
There was a point about 7 years ago when the overton window shifted to "speak english to strangers first" because of a large influx of foreigners who did not know the language. Since then I've met foreigners who have better Maltese than some natives.
Older folks & geriatrics will sometimes be surprised when they assume someone is foreign and they turn out to be Maltese. "int Malti??" is a statement I get often because I don't look Mediterranean despite being born here.
How is "Marsaxlokk" really pronounced? I've heard that word a few times, but never from a native. Google translate can't help me here, as it doesn't seem to have Maltese text-to-speech.
Read with English pronunciation, closest would be mar-sa-shlock.
From my experience it will be understood by locals when pronounced like that.
What is the name of Maltese in Maltese? Like “el español” in Spanish, it’s neat to know what languages call themselves
[delayed]
Wikipedia says it's "Malti"
Il-Malti to be precise. Il- means "the" and changes its meaning to that of the language. Malti alone would mean a Maltese person.
Source: I'm also Maltese.
The "Il" in Il-Malti is like "al" in Arabic, which Maltese is closely related to as was pointed out above.
Arabic (language): al-‘arabiyyah (الْعَرَبِيَّة).
'ish' is a pretty universal english suffix. So Spanish is just "españ-ish".
Is there any dialect of Arabic which you can understand without too much effort?
How much do you consider Maltese its own language (as opposed to a dialect of Arabic)?
From what I have heard, Lebanese Arabic is the closest, and still pretty far. Passable conversation is possible.
Maltese is definitely its own language. Arabic roots are there (theres a Semitic joke in there ) but it isn't arabic anymore. Its written left to right with a variant of the english alphabet.
I know that the reverse understanding isn't too bad from chatting with a Saudi-born member of staff on holiday in Malta.
I don't think anyone would seriously consider it a dialect of Arabic though with its completely different alphabet and half the vocabulary and morphology coming from Italian languages/dialects, even if Malta hadn't spent the best part of a millennium trying very hard not to become part of the Arab world
Can you communicate with Maltese dogs more effectively?
Only if we have a few Maltesers first
> Greek being the only Hellenic one
Are there really any other Hellenic languages besides Greek?
Seems like the model isn't limited to those though, from the paper:
> as well as some additional relevant languages (Arabic, Catalan, Chinese, Galician, Hindi, Japanese, Korean, Norwegian, Russian, Turkish, and Ukrainian).
https://arxiv.org/pdf/2409.16235
The paper also goes into detail on training set sources, which I feel like a curation thereof might be considered the main contribution of this publication?
Lithuanian and Latvian are Baltic languages. Nothing to do with Slavic...
https://en.wikipedia.org/wiki/Balto-Slavic_languages
See the section "historical dispute".
I think some people get touchy about them being lumped together if their last period of commonality (per the article) was 1400 BCE. For comparison, I believe all the Slavic languages were mutually intelligible around 1200 AD. But much more recently than this, in the last few centuries, there have been notable attempts by east slavs to absorb the Baltic language cultures and deny them.
I doubt that South Slavic and West/East Slavic were mutually intelligible at 1200 AD.
I doubt West and East Slavic were. But inside those geographic groups they probably were (Czech and Polish AFAIR were around that time).
Depends on your standards, too. Even today, any pair of slavic speakers should have a head start in understanding each other. Put them next to each other for a month and they should be talking, at least about basic everyday things.
Balto-Slavic branch divides into Baltic and Slavic language groups so nothing wrong here
It is just one of the theories, there is no clear evidence to suggest that Baltic and Slavic were the same language thousands of years ago.
Well there is if you go far enough. It's just the question when did they split off from each other. However there is no question that Baltic and Slavic are more closely related to each other than any other non extinct Indo-European languages.
The fact they they are the closest surviving relatives on it own doesn't mean it makes sense to group them together (i.e. Italo-Celtic is also a theorized subgroup in a similar way but nobody is disputing that Celtic and Italic languages evolved into distinct groups).
Then there is a huge amount of missing links and unknown unknowns. e.g. Thracian and Dacian probably were also pretty close to Baltic or Slavic (maybe even closer to Baltic than Slavic is but we don't know enough about them to make any conclusive claims at all... but we at least know these languages existed)
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.
I regret being that loose with the designation :), Romanian and Hungarian are valid counter arguments.
In my mind, I was thinking of the belt of countries between Russia and Central Europe, starting from the Baltics down to the Balkan (excluding Greece).
Even by your definition, I can count at least seven countries where the official language is not Slavic. And that's not even including all the Altaic, Romance and other assortment of regional languages, many of which have some sort of official status.
Albania is not "East Europe", but South East. Same as Greece.
That's just your opinion, and the UN would disagree: https://www.un.org/dgacm/en/content/regional-groups#:~:text=...
Some of my fellow Romanians will also claim they're Central European, but in my mind, all the ones I listed are Eastern European countries. I'd even include Turkey and Kazakhstan in there, part of the latter is to the West of the Urals, which is what we normally consider the border between Europe and Asia.
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.
Hungarian too, although there’s a question about whether Hungary is Eastern or Central Europe.
“There’s a question” implies that there is a ground truth that might be discovered to resolve this rather than simply a clash of different purely arbitrary definitions of the same terms.
Ah, yes, how could I forget! As a side note, though also Finno-Ugric then similarity in sound and appearance from Finnish or Estonian at least appears very far.
The Visegrad 4 (Poland, Czechia, Slovakia, Hungary)are generally taken to be "Central European". The strict East/West division is largely a product of the Cold War and the Iron Curtain.
> most of Eastern Europe are Balto-Slavic
and
> only Estonian is not a Slavic language.
So following this logic saying "in Eastern Europe, only Estonian is not a Baltic language" would make as much sense?
Plenty of wrong here, considering Lithuanian and Latvian are utterly unintelligible to slavs, save for loanwords, but Slavic languages between themselves retain some level of intelligibility, which even spawned two competing constructed languages.
I was thinking about separating the two groups when I was writing this but was afraid of getting too verbose, though in retrospect that probably would have made more sense regardless of the historical lineage. My apologies if this came off as inconsiderate.
I updated my original comment, and learned a good amount about that dispute as a result, so thanks for calling it out.
I find it interesting that Norwegian isn't on the list.
I have often joked that Norwegian is just a dialect of Swedish, but I never expected to get official validation like this!
Norwegian is not on this list, because in fact no country with Norwegian as their national language is part of the European Union at the time of writing.
Norway isn't in EU, though.
Norway isn't in the EU.
Flemish? I remember watching a TV show in Flemish (Hotel Beau Séjour[0]), so it's prevalent enough to invest that kind of money into.
What about Basque? Is that too controversial?
[0] https://en.wikipedia.org/wiki/Hotel_Beau_Séjour
Flemish is more of a political construct than linguistic, it's a grouping of belgian-dutch the coastal, brabant and limburg language groups with each having their own regional dialects.
It's more than political. In speaking Flemish is to Dutch as UK English is to US English. In writing however there is no difference in spelling, but there is a difference in word choice.
Now, being from Belgium, even within that small part of the country where everybody is supposed to speak Dutch, I genuinely don't understand people from near the coast, which was about 150 miles from where I used to live.
Basque is not controversial, but spoken just by very little people.
Not sure that should be the qualifier, there might be more people able to speak Basque in the world than Danish, doesn't stop Danish from being well supported.
Quick google points to about 1M Basque speakers in the EU against 5-6M Danish speakers, there's also the fact that Basque is not the only official language in the country it belongs to, and that it's in fact not spoken in the vast majority of the country.
From https://european-union.europa.eu/principles-countries-histor... we can find an excerpt relating to the policy and its purpose:
>One of the EU’s founding principles is multilingualism.
>This policy aims to:
>communicating with its citizens in their own languages
>protecting Europe’s rich linguistic diversity
>promoting language learning in Europe
With this in mind, the first intention fails by an enormous margin, given that 95%+ of Spain doesn't speak an iota of Basque, the second is met handily, given the long history of the language, and I'm not sure what to think about the third, any language whatsoever would serve that purpose.
like British English vs US English, Flemish is a dialect of dutch
Ah. That makes sense.
It's all Greek, to me...
I think those 24 languages reflect all the languages that are official languages at country level.
So for instance, Basque is not an official language of any country (only French in France and Spanish/Castilian in Spain). Belgium's official languages are French, Dutch, and German, "Flemish" is only a local variant of Dutch (Belgian French is also only a local variant of French).
Official is a weird concept though. Turns out Dutch law never really bothered to define an official language, Dutch simply is the de facto standard and is required for a lot of things making it effectively the standard. This makes Dutch Sign Language the only language officially recognised by law. An attempt to recognise Frysian and Dutch as official languages in the constitution failed.
Sweden didn't have an "official" language before the Language Law of 2009. Five minority languages (Finnish, Meänkieli, Romani, Sámi, Yiddish) were officially recognized as such since 1999.
Basque is an official language and declared as such in the Spanish constitution however restricted only to the regions that decide to apply it (Basque Country and Navarra).
If we want to go all legal, I believe that Spanish/Castilian is the only official language of the State, so at country level, with the other "Spanish languages" only official in their respective areas:
Section 3
(1) Castilian is the official Spanish language of the State. All Spaniards have the duty to know it and the right to use it.
(2) The other Spanish languages shall also be official in the respective Autonomous Communities in accordance with their Statutes.
(3) The richness of the different linguistic modalities of Spain is a cultural heritage which shall be specially respected and protected. [1]
[1] https://www.senado.es/web/conocersenado/normas/constitucion/...
Thanks. That makes sense.
In the US, people will resort to fisticuffs, over variants of Spanish. I usually translate into Castilian Spanish, because that seems to be the equivalent of "Vanilla" Spanish. No one is really happy (except the Spaniards), but I'm not accused of favoritism.
In Greek we call our language Hellenic, and our country Hellas. "Greek" / "Greece" don't exist in the Hellenic language.
Yes it does, it was a greek colony off the southern coast of Italy, which were the primary greek connection to the romans which how the name stuck.
> Γραικοί, Graikoí were an ancient Hellenic tribe
https://en.wikipedia.org/wiki/Graecians
Maltese is derived from dialectical arabic
Is Ireland the only country to bring in two languages, Irish/Gaelic and English? Is English an official language of any other EU countries?
English is an official EU language because Regulation 1 Article 1 says so [0] and hasn’t been changed. In practice, English is the most widely used language in EU institutions, so it would be have been silly to remove it after Brexit.
[0] https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:01...
It's a national language in Malta, making it a popular destination for "language weeks" in European schools, where English is usually a main subject.
English at this point has stopped culturally belonging to the United Kingdom and whilst one can discus it's not so very moral way of getting there, it's become the bridge language for people of different languages to communicate in, further solidified by the internet.
That said, whenever there is a language selection UI (e.g. at banking machines or institutional websites) in wider Europe that uses flags to represent languages -- probably not a good idea to start with, but very common -- the Irish tricolour should be used to indicate English rather than the UK or USA flags. (although cf Airteagal 8 of Bunreacht na hÉireann).
I believe Malta has English as an official language.
PS: Gaelic is a more general term for Irish and Scottish. Ireland brings specifically Irish(Gaeilge in Irish) language.
Malta has Maltese and English as official languages. I don't know what they bring to the EU list of official languages.
AFAIK Ireland only listed Gaelic as their official language with UK having English. That caused a bit of a problem during Brexit since technically English wasn't officially an EU language anymore. I guess they resolved it somehow.
From other comments, it seems many people don't realize that there are 11 more languages than these 24 official (this is mentioned in the paper):
Arabic, Catalan, Chinese, Galician, Hindi, Japanese, Korean, Norwegian, Russian, Turkish, and Ukrainian.
+1
No Luxembourgish?
Turkish?
Is official in Northern Cyprus. But as I understand it while the whole island of Cyprus is in the EU, the state of Northern Cyprus isn’t.
Norwegian is also included, based on the model card: https://huggingface.co/utter-project/EuroLLM-9B
Actually nuts to me the degree to which European policymakers do not even begin to understand how to kickstart technologically-intensive industry. Anyone who has seen close-up the results of a "pick the winners" grant-style approach to innovation knows what will go wrong here.
Also funny to read this narrative of how access to the European 'supercomputer' cluster is going. https://x.com/levelsio/status/1981485945745788969
While grant process in EU isn't fun, I think Levels has bit of an ego issues. He mentioned that if he had issues like that on eg X, he would see Elon himself in the replies.
While he is great at converting his influencer status to income in his micro-SaaS projects, I don't think running ad-fueled browser games on state-sponsored super computer should be really aim of these grant programs.
He is 100% right on this one. From personal experience trying to figure out EU. Lawyer bureaucrats manage funds behind red tape clearly meant to be for their pals.
All these while the EU is running out of funds and in a process of de-industrialization. There should be an independent corruption investigation on Brussels.
I took part in application for EU grants a few times and our company group did it many times over the years.
It's bureaucracy, often bordering with stupidity. You may need advisors to navigate all their forms & processes. But it certainly isn't "pals-only" type of deal.
On the other hand - is it harder than getting VC funding? For seasoned founder with reputation - probably. For fresh startup - probably not.
Reviewer (Scientific Expert) for the EU (since 2009) here.
The probability of getting a Horizon Europe grant allegedly (not official stats) is about 8.5% according to some friends, which may seem low. You need to write 70 pages following a Word template and the key goal is to cover answers to a large number of questions. Each proposal gets various grades across a range of dimensions, which get added up and if you obtain at least 13 out of a possible 15 points, you are eligible to get funded, read: "You will get funded if there is enough money." Often, there are several proposals that justly achieve 15/15, and because of that, many prosals that have 14 points and all proposals that have less may not get funded, simply because there just is not enough total funding available to fund all the technically eligible proposals. Having judged many proposals in AI / ML / search / "big data" / language technology etc. I recommend optimizing recall, i.e. aspiring completeness.
The application process is not easy, but you can get help: there are support agency in each member country, free online Webinars to help, hotline help desks as well as an ecosystem of paid consultants that typically charge about 3k€ to vet a proposal for you if you need that kind of service (I never used it).
The process is neutral and conducted professionally and with external oversight (consultants are hired as "rapporteurs" that report on process/procedural integrity in additional to the actual reviewers). I value the research officers of the EC as people of high competency, integrity and motivation (research money is tax payers money so it should be spent carefully).
In comparison, VC (and even more so business angel) funding is achievable with much less formal apparatus, often a short business plan and a convincing slide deck and demo can get people to a partner meeting if the time is right. But the criteria and process are much different, and ideas ready for public research grants are typically too early for VCs (but the EC wants to foster the creation VC-funded startups resulting from the disseminated research).
> For fresh startup - probably not.
highly doubt, the whole thing about the success of the US west coast is that they are&were willing to fund unproven upstarts.
Right but if we do this with public funds then the narrative shifts to "OMG the EU is so corrupt and stupid, looking they're pouring taxpayer dollars into unproven stuff! They're deindustrializing!!"
The point being that, as soon as public dollars are on the table, people expect perfection. Anything less is waste, fraud, and abuse.
There's literally no winning. Want to make sure the money is allocated right? Bureaucracy. Want to not do that? Waste, fraud, and abuse.
The winning move is that governments should do government stuff and private capital should do private capital stuff. Startups belong to the latter.
> Startups belong to the latter.
Except that Apple, Intel, Tesla, etc have all received US government investment [1]. TSMC is a product of the Taiwanese state! Government investment can be done well, and seeds excellent companies.
[1]: https://www.sba.gov/blog/2024/2024-02/white-house-sba-announ...
That's exactly the problem in Europe though. It's quite the opposite here.
Someone told me I needed to hire some expensive law firm in Brussels. See:
https://www.politico.eu/article/ombudsman-slams-commission-f...
Of course there is red tape. EU funding comes from taxpayer money and we want it to be spent wisely. The red tape is precisely to prevent it from being funneled to pals. EU has funded quite a few free software projects so it's not like the red tape is an insurmountable burden: https://www.ri.se/en/news/blog/europes-digital-future-spells...
I'd also say that their grants aren't unusually burdensome and grantmaking is arms length compared with a lot of other bodies.
Yes, some of the questions are weird, but I'd really rather write a bit confirming that the AI system being developed isn't going to be racist or Skynet than jump through some other hoops that exist (and that absolutely includes VC due diligence). The actual biggest issue with European funds is they get way more competent applications than they can fund anyway.
I'm actually no fan of his, so that's fine. That said, I went to the actual website he was talking about (I'm also an EU citizen) and in this case it is exactly as described and bordering comical.
It's not even close to how he described it.
There’s a screen recording at the bottom
I have the same answer as here: https://news.ycombinator.com/item?id=45735738
> 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 adding regulation.
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.
Ironically, China has in some ways a bigger regulatory burden when it comes to software, as there if the government doesn't approve the business is dead in the water. I doubt that Klarna would've gotten off the ground there, for one, I could see them being shut down much earlier there. In the EU only now very slowly are some governments even starting to talk about some weak measures around their business model. But I've never, not once in my life, heard "Chinese software companies can't get off the ground due to the regulatory burden".
The same people who clamor about the EU regulations are the ones who hate on the EU for their protectionist measures against US tech. Yet another bout of irony here - China's software industry has flourished exactly thanks to 10 times stronger protectionist measures against US tech. So has Korea's, and their protectionism has never even been anywhere on the China level, more inbetween EU and China. No, if there's anything that would help, it's much more tech protectionism in the EU.
Pieter Levels is at the end of the day an influencer, not a serious founder.
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.
Where in Europe and where in the US? You probably started one in the easiest US state to do so. Did you try starting one in the easiest EU state? Otherwise we already can't take things very seriously.
Secondly, what's easier besides VC funding? If it's VC funding, the disparity there has nothing to do with regulations - guess how much VC funding the non-EU rest of the world gets.
What would you want to see changed to consider coming back?
> but I started my company in the US because everything is so much easier here
Which part is easier? That you have 50 different states with slightly varying laws to consider (e.g. Californian Data protection)? That you have a byzantine system of "benefits" to choose and manage?
And compared to where? Germany or Estonia or Sweden or Spain? The complexities will vary wildly depending on the country (kind of like in the US, where lots of companies pick the state to base themselves in based on the combination of favourable laws and precedents and taxes).
"That you have 50 different states with slightly varying laws to consider (e.g. Californian Data protection)?"
there are certain sentences you can just tell would never be written by an American lol
Got me, I'm not American, but isn't it true?
California Consumer Privacy Act is a thing you need to take into account for Californian customers.
Illinois has a Biometric Privacy Act.
And who knows what Wyoming or South Dakota or Oregon have that you might take into account if your business falls under any of them.
we might be somewhat trending in this direction, but the reality is largely that the US states are pretty identical and have very similar laws on the books. the federal government is in charge of commerce usually.
most laws like CCPA also have some threshold where you already need to be pretty successful for it to apply to you.
for some select industries (biometrics & healthcare), yes you have a patchwork of laws.
> 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.
Yep, VCs don't exist here. Plus the absurd starting costs, it's like what, 20k to set up a GmbH?
2.5k EUR in starting capital, and two founders to start a a limited liability company (AB) in Sweden, and a 240 EUR processing fee: https://verksamt.se/starta-foretag/valj-foretagsform/aktiebo...
And you register online.
Depends on the country.
Opening a company in Estonia is very cheap but in Spain the manager/CEO needs to be an "autónomo" (like a self-employed tax status). This costs thousands of Euros per year. Something like 2,400-30,000 Euros per year, every year, forever.
And that's probably one of the big obstacles in the EU: there's no common ground for these things. At least this will hopefully be addressed: https://www.reuters.com/business/eu-propose-uniform-rules-st...
> 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 competitive wages. There's a real difficulty in allocating capital to talented humans.
And, at least in Southern Europe, the income tax schedule is so aggressive it's hard to justify continuing working in many of these countries if you are highly talented.
Like, if you can tell me what the induced operator norm from l_2 -> l_2 is - probably you should come to the US and work at a biglab and make bank. What can you do in Portugal, Italy, Spain, etc.??
> Pieter Levels is at the end of the day an influencer, not a serious founder.
Sure, agreed.
I think it is a complete misreading to point to protectionism as the reason for Chinese success, but having a big unified domestic market for consumers along with massive saving rates and capital controls probably does help.
One fairly large factor is that even though English is much more common today, you just can't operate (depending on the product of course) in many countries without having customer support, documentation etc in the local language.
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.
Are you saying that the other 199 non-US countries in the world all have a business unfriendly environment, since every one of them besides China has practically the same amount of software VC funding compared to the US?
All of these purported EU-specific reasons completely ignore that things are the same elsewhere. It's the US that is the outlier.
> I think it is a complete misreading to point to protectionism as the reason for Chinese success, but having a big unified domestic market for consumers along with massive saving rates and capital controls probably does help.
Capital controls are protectionist measures, but anyway, no.
> Okay, what is the limiting factor?
Let's look at which countries have a significant local software industry compared to population size.
- China
- US
- Korea
- You can argue for Japan and India but that's already starting to stretch.
- Yup, effectively no where else. Even in an "out of the way" place like Myanmar everyone uses Meta, with a nice little genocide to show for it. Sure, in Vietnam they use Zalo, and other places have a few other local players. But most of the famous US tech apps are dominant.
Is the EU the outlier here? No. Everywhere else US tech dominates. Meta, Netflix, Apple, Google, Uber, Spotify, Microsoft, Match Group, Paypal, Amazon, and on and on. They don't just dominate the EU, they dominate the world.
Except for the countries I named above, where at least some of the markets that US big tech competes in, instead have bigger local players. And even there, guess what?
Their market share is almost 1:1 linearly correlated to the degree of protectionism in those countries, all the way from China, then Korea, then India/Japan, and then everywhere else! Who woulda thought!
Why does Korea have much less US tech dominance than, say, Germany? Despite German companies theoretically having a big advantage: the German public is 100x more privacy conscious than the Korean one, and much less trusting of US companies.
I can tell you that it's not less regulations; Korea's GDPR is much more onerous than the EU's and so are investment regulations. On every single regulatory aspect, German software startups have it easier. But they were never protected. US tech was allowed to waltz in, dump their products - that's what they did, it's hilarious how now China "dumping" EVs and solar is suddenly an issue when it's exactly the strategy that US tech continues to this day; the AI companies are doing it right now! And the Korean companies were protected. Both by the rules burden, that local companies had to deal with too, along with intentional protectionism.
When it comes to solar and EVs, we all understand that a foreign country dumping their goods kills local industry. It's the exact same with software.
But then half of HN has millions on the bank exactly thanks to the above - this is where all those fat SV salaries have come from - so I do get the lack of desire to understand it.
> Their market share is almost 1:1 linearly correlated to the degree of protectionism in those countries
Seems like you actually believe this. I think our starting points on reality are different enough that we are not going to have a productive conversation, I wish you and other Europeans the best of luck in your protectionism-led growth strategy. Make sure to not discuss it with any pesky macroeconomists who might lead you astray. take care
I've provided very specific cases that directly support this, you've so far provided nothing. This is a really poor comment.
Spotify is Swedish. Uber is irrelevant in many places in the EU due to protectionism.
Spotify is not a US company.
> Okay, what is the limiting factor
A few.
A big part is that the EU is a collection of countries that (with very few exceptions) have different languages and laws. For a company to serve Spain and France, for instance, it would need to translate everything, hire local lawyers and customer support agents. Considering the much smaller size of the countries (biggest one is 70 million vs 330 million in the US), the opportunity for "unlimited" growth is limited.
This also rebounds in the fact that when an American company makes it big, they have the resources to flood other EU markets and be cheaper/better than the local competition due to economies of scale and money based on their big successful US market. A French company making it big is still small compared to a US equivalent.
Then, there's the capital markets, no denying that. The money being thrown around the US is like nowhere else on the planet. Some of it definitely a bubble / unrealistic, but that doesn't matter. But in part it's because of the size of the total potential market that this is justified.
Education / national mythology also plays a part, I think (this is pure conjecture now). In the US, the "American Dream", "everyone can make it" etc is heavily ingrained. It propagates through the world with the help of Hollywood and other American cultural exports. In most EU countries, there isn't such a heavy emphasis on independence and "pulling yourself up by your bootstraps". "Hustle culture" isn't a thing. So for most people, it isn't something that comes naturally to them to start a company and work 100 hour weeks to be big and rich and successful and famous.
That's not to say there aren't such people, I went to 42 and have been to Station F and know some people in that universe. A decent proportion of my classmates wanted to make their startup and make it big, and some did end up starting their own companies.
> This also rebounds in the fact that when an American company makes it big, they have the resources to flood other EU markets and be cheaper/better than the local competition due to economies of scale and money based on their big successful US market. A French company making it big is still small compared to a US equivalent.
Ding ding ding! When China does it with solar and EVs we call it "dumping". When Uber, OpenAI and Anthropic do it, that term is never ever used. VC funded US techs dumps harder than any Chinese industry ever has.
Tbh, a lot of EU protectionism vs. US tech seems not to keep the competition out. In fact, with the amount of free press US startups get and the size of their coffers, they can simply roll over the local competition in EU markets most of the time.
What it's terribly good at is adding burdens that the US giants don't face early on, slowing down the early growth between 28 fragmented markets. I don't know specifically about how China works, but the question is proving product-market fit, and for that, you need a lot of users fast.
In the EU, it's a different battle country to country as the media environment, the markets, the regulation etc. are all fractured.
What are the effects of pick the winner strategy? Sounds intriguing
Is the point of these policies to pick winners? Or to upskill the creators and stimulate the economy by giving possible entrepreneurs experience Europeans can't get in big tech?
In the US, some ex-Googler might found a startup. Europe doesn't have the equivalent of FAANG. (Europe-wide companies are not quite as easy as US-wide)
Even if the super computer itself "fails", is the goal actually the secondary impacts to the economy?
(And in the US, we do our own fair share of picking winners / losers, especially in the current regime)
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.
I don't think what you're saying is inconsistent with what I'm saying. I think you are making a big deal out of the difference between state investment funds and subsidized GPUs but I think they basically work by similar mechanisms.
> 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.
Does government offering matched funding to investors actually help startups who are struggling to find (any) funding? If a startup can't find (any) funding, matching is irrelevant.
> 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.
Umm. I'm not really convinced that the political elites in Europe understand how to do any of this stuff well.
See also: https://www.eib.org/en/publications/online/all/the-scale-up-...
This guy spreads FUD about the "unelected commission". What a loon.
Levels is engagement farming. Instead of uncritically reposting him you could've gone ahead and read what the cluster is for: https://x.com/dmitriid/status/1982927767286231403
Cluster: for public benefit, cutting edge research in biotech, medical, robotics.
Levels: I want to create AI photos of people for my AI Slop startup
> Cluster: for public benefit, cutting edge research in biotech, medical, robotics.
That's not what the quoted paragraph says and you can read the whole release if you want: https://ec.europa.eu/commission/presscorner/detail/en/ip_25_...
I literally quoted the paragraph from this link in the tweet I provided: Edit: lol, I didn't, I quoted it from a policy document, not from press release. However, my point stands:
--- start quote ---
Apply AI Strategy
The Apply AI Strategy aims to harness AI's transformative potential by driving adoption of AI across strategic and public sectors including healthcare, pharmaceuticals, energy, mobility, manufacturing, construction, agri-food, defence, communications and culture. It will also support small and medium-sized enterprises (SMEs) with their specific needs and help Industries integrate AI into their operations.
--- end quote ---
I also quoted a paragraph from a document I will find when I'm not on mobile.
Levels literally wants to train AI Slop: https://x.com/levelsio/status/1981499900266193028
--- start quote ---
Train a foundational model for AI photos of people
--- end quote ---
Seems like your quote was very misleading to me, so no your point doesn't stand.
> Seems like your quote was very misleading to me, so no your point doesn't stand
My quote: Cluster: for public benefit, cutting edge research in biotech, medical, robotics.
Literal quote from your link: The Apply AI Strategy aims to harness AI's transformative potential by driving adoption of AI across strategic and public sectors including healthcare, pharmaceuticals, energy, mobility, manufacturing, construction, agri-food, defence, communications and culture.
You: your quote was misleading.
I'm sorry, I don't have the time or the patience with willfully ignorant and blind people getting their interpretations from AI slop engagement farmers.
Adieu
Unfortunately the AI Slop is probably the most effective way to fund AI research right now
But the point here isn't to fund AI research, it is to use AI to benefit concrete fields.
By funding AI slop, you're funding AI slop, not AI research, or, quote, "drive adoption of AI across strategic and public sectors including healthcare, pharmaceuticals, energy, mobility, manufacturing, construction, agri-food, defence, communications and culture"
EU grifting is so much worse than even the most brazen Trumpian crypto pump n' dump.
Geniunely repugnant. Atleast the Trump admin has the decency to pump everyones 401k...
I'm trying to figure out why it bothers me so much. I think its because the EU are such unbelievable losers in everything they do. they can't even grift, thats how useless they are. they can't even steal properly. its so undignified, and offensive to the senses.
Wouldn't go that far. EU policymakers have good intentions, I believe - but ultimately are products of their environment and cultural inclination.
The EU is such a bizarre place because they treat capital and entrepreneurs with such massive distrust, but never really bothered getting rid of the quasi-static entrenched hierarchies from feudalism? Like I'll go to the UK or France and there will just be massive swathes of land owned by the nobility or 'former' nobility? Maybe start there but let your high-value human capital earn a good wage?
You are wrong in that you think the hierarchies stem specifically from feudalism, but you are absolutely correct in that these hierarchies exist and are deeply entrenched. Sweden and Germany have one of the lowest percentages of self-made vs. inherited fortunes in the western world. Actually some tax policies in the US enable much more upward mobility, such as real estate taxation and 401k-like vehicles.
> France and there will just be massive swathes of land owned by the nobility or 'former' nobility
Yeah, no, this isn't even remotely true.
will cede that, you're right for France.
Title is missing "(2024)". The 9B model was released last december[0].
0: https://sites.google.com/view/eurollm/home
>The EuroLLM Team brings together some of the brightest minds in AI including Unbabel, Instituto Tecnico Lisbon, the University of Edinburgh, Instituto de Telecommunicacoes, Université Paris-Saclay, Aveni, Sorbonne University, Naver Labs, and the University of Amsterdam.
>Europe is the only continent in the world to have a large public network of supercomputers that are managed by the EuroHPC Joint Undertaking (EuroHPC JU). As soon as we received the EuroHPC JU access to the supercomputer, we were ready to roll up our sleeves and get to work. We developed the small model right away and in less than 6 months the second model was ready.
[1] https://www.eurohpc-ju.europa.eu/eurohpc-success-story-speak...
Repurposing some of that physics sim compute
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.
No, that's not how training works. It's not just about having an example in a given language, but also how many examples and the ratio of examples compared to other languages. English hugely eclipses any other language on most US models and that's why performance on other languages is subpar compared to performance on english.
Ratio/quantity is important, but quality is even more so.
In recent LLMs, filtered internet text is at the low end of the quality spectrum. The higher end is curated scientific papers, synthetic and rephrased text, RLHF conversations, reasoning CoTs, etc. English/Chinese/Python/JavaScript dominate here.
The issue is that when there's a difference in training data quality between languages, LLMs likely associate that difference with the languages if not explicitly compensated for.
IMO it would be far more impactful to generate and publish high-quality data for minority languages for current model trainers, than to train new models that are simply enriched with a higher percentage of low-quality internet scrapings for the languages.
I have never noticed any major difference in performance of ChatGPT between English and Spanish. The truth is that as long as the amount of training data of a given language is above some threshold, knowledge transfers between languages.
There's actually a research showing that llms are more accurate when questions are in Polish: https://arxiv.org/pdf/2503.01996
> because they are trained on multilingual data
But they were not trained on government-sanctioned homegrown EU data.
Who in their right mind would use this?
I'd use a model trained on a targeted and curated data set over one trained on all the crap on the internet any day.
I keep hearing that LLMs are trained on "Internet crap" but is it true? For instance we know from Anthropic copyright case that they scanned millions of books to make a training set. They certainly use Internet content for training but I'm sure it's curated to a large degree. They don't just scrap random pages and feed into LLM.
> I'm sure it's curated to a large degree. They don't just scrap random pages and feed into LLM.
How would they curate it on that scale? Does page ranking (popularity) produce interesting pages for this purpose? I'm skeptical.
The entirety of the internet vs government-sanctioned homegrown EU data.
"But they were not trained on government-sanctioned homegrown EU data."
ok what are you implying on this
> But they were not trained on government-sanctioned homegrown EU data.
If none of the LLM makers used the very big corpus of EU multilingual data I have an EU regulation bridge to sell it to you
European governments have huge collections of digitalised books, research, public data.
But also European culture could maybe make a difference? You can already see big differences between Grok and ChatGPT in terms of values.
If it's publicly available data, books and research, I can assure you the big models have already all been trained on it.
European culture is already embedded in all the models, unless the people involved in this project have some hidden trove of private data that they're training on which diverges drastically from things Europeans have published publicly (I'm 99.9% positive they don't...especially given Europe's alarmist attitude around anything related to data).
I think people don't understand a huge percentage of the employees at OpenAI, Anthropic, etc. are non-US born.
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.
Meh, it depends a lot on the dataset, which are heavily skewed towards the main languages. For example they almost always confuse Czech and Slovak and often swap one for the other in middle of chats
But the only way to unskew it is to remove main language data because there isn't really any to add, no?
You can also correctly bias your sampling so that when selecting new training instances each language is chosen equally. Generally the diversity of data is good, unless that data is "wrong" which, ironically, is probably most of the internet, but I digress.
Aren't they about as different as American English and British English?
The difference ia larger than let's say just a "dialect". They really are different languages, even though we generally understand each other quite well (younger generations less so). I've heard it's about as different as e. g. Danish and Swedish - not sure if that comparison is helpful.
Nope. Capability begins to degrade once you move away from english.
Plus all your T&S/AI Safety is not solved with translation, you need lexicons and data sets of examples.
Like, people use someone in Malaysia, to label the Arabic spoken by someone playing a video game in Doha - the cultural context is missing.
The best proxy to show the degree of lopsidedness was from this : https://cdt.org/insights/lost-in-translation-large-language-...
Which in turn had to base it on this: https://stats.aclrollingreview.org/submissions/linguistic-di...
From what I am aware of, LLM capability degrades once you move out of English, and many nation states are either building, or considering the option of building their own LLMs.
The EuroLLM-9B model release is from Dec/2024, and scores just above random chance for benchmarks like MMLU-Pro (17.6%, random chance is 10%).
Comparison with similar EU models + 600 other highlights:
https://lifearchitect.ai/models-table/
I was thinking the same, why are so many superior models coming from only countries like US and China. And why are European countries not in the list other than France with Mistral. Why are so few companies in India, Japan, South Korea even close to a promising new model like what Chinese companies did ?
"Why" is a fair question but are you surprised? Europe is consistently behind in tech.
Europe has about 1.3 times the population of the USA and about 75% of the GDP yet EU tech output is a very small percentage of US tech output. We are not talking about 70, 50, 30, or even 20%. It's a drop in the bucket.
>The seven largest U.S. tech companies, Alphabet (Google), Amazon, Apple, Meta, Microsoft, Nvidia, and Tesla, are 20 times bigger than Europe’s seven largest, and generate 10 times more revenue.
https://eqtgroup.com/thinq/technology/why-is-europes-tech-in...
"Why" is a good question, but I definitely wouldnt expect significant competition in LLMs from Europe based on the giant tech disparity. Having 1 non-cutting edge model that isn't really competitive is pretty much what I would expect.
> The seven largest U.S. tech companies (...) are 20 times bigger than Europe’s seven largest, and generate 10 times more revenue.
I'm going to guess that this part is intentional. Europe tends to be more aggressive in enforcing antitrust laws. Economically, Europe's goal isn't to have the biggest companies but to have more smaller companies.
So you're not going to get companies like Google, but you will get companies like Proton, Spotify, Tuta, Hetzner, Mistral, Threema, Filen, Babbel, Nextcloud, CryptPad, DeepL, Vivaldi, and so on.
>I'm going to guess that this part is intentional. Europe tends to be more aggressive in enforcing antitrust laws. Economically, Europe's goal isn't to have the biggest companies but to have more smaller companies.
So is your hypothesis that the total market cap of EU tech companies is something like 50,60,70, etc. % of total US tech marketcap? Something significantly different than the ~10% implied by that figure (largest us companies 10x largest EU companies). And it's just more broadly distributed?
Hard to find data on this but this is showing EU tech market cap at 3.2T. https://www.stateofeuropeantech.com/chapters/outcomes
Whereas this is saying the US "megacaps" ($200B+) are at 21T. https://www.cnbc.com/2025/09/05/tech-megacaps-worth-market-c...
Which puts the entire EU tech market at 15% of the US megacaps. Not even the entire market.
European companies are smaller on average and less likely to go public in general, so market cap comparisons don’t show the whole picture. Growing big is less often seen as a goal than in the US. “Megacaps” aren’t necessarily considered a healthy thing to have.
Yes, and this all but guarantees that Europe will stay behind USA and China in their technology capabilities.
What are these capabilities?
I don't see any sense in which the EU has fewer capabilities. It has, say, a smaller number of businesses with smaller market dominance.
It isnt clear to me what capability the EU would gain by having a monopolist social network, a monopolist search engine, a monopolist advertising trader
Europe has all of those things, they just come from the US.
Also, commercial software is consistently behind from open source.
I only use open source LLMs for writing (Qwen 32b from Groq) and open source editor of course, Emacs.
If some people can write better using commercial LLMs (and commercial editors), by all means, but they put themselves at a disadvantage.
Next step for me, is to use something open source for translation, I use Claude for the moment, and open source for programming, I use GPT curently. In less than a year I will find a satisfying solution to both of these problems. I haven't looked deep enough.
As a European citizen I think it boils down to access to the capital. EU/EEA is not a country and the market is sort of fragmented. The big players are UK, France, Germany, everyone else does not have the same access to money as say in the US. Folks want to do it but there is a glass ceiling. Hence you have these collabs among large institutions to tap into funds such as from Horizon which are academic in nature and do not translate well into products.
Because training frontier model is expensive and only US and China have capital structure to raise tens of billions of dollars to do it.
You can easily fit below 10 billion for the whole datacenter, then you only pay for electricity + maintenance + staff. 100k GPUs cost a few billion USD, that's more than enough to train frontier models, run experiments, and serve models in the EU to start. Look at what xAI did and how much it cost them and it's more expensive to do in US than in EU.
being able to train new frontier models is the new equivalent to nuclear capabilities.
i predict at some point countries will get CIA'ed when they publish plans to build a large data center.
Similar to the time when they got CIA'ed when announcing plans for new nuclear plants.
They are already CIA'ed on a regular basis for much less than that.
EU made a >900 page law about AI and patted themselves on the back for being "the first to regulate AI" (which was not even true, China had an AI law before and it's two pages long).
This cannot be stressed enough. In my experience working in multiple tech startups in Germany, the power compliance, legal and all other 2nd line has over engineering is quite immense. Most of the time they act as a hindrance for innovation rather than a supporting factor.
This AI law is a clear example of that. Pencil pushers creating more obstacles for the sake of creating more obstacles rather than actually taking a pragmatic approach.
It's strange, my real life experience is very different than yours. Unless you're training AI to do something shady, it's really no bother at all. In fact, most of what the AI Act requires, you have to do anyway for a good model card.
Because the value of these models is (actually) yet to be proven. Why saturate the market with something that we already have at least one of and others are selling as a service? No model provider (including the "big ones" like OpenAI) has been able to produce a viable business case. They're all literally running on government deals and investor money.
Does it even make sense? Just use the American or Chinese ones, adjust As needed. Where’s the point in spending millions to build The same thing or worse
Now that the big bets have been made, who wants to try to compete with them?
From the EuroLLM-9B page on hugginface;
>You need to agree to share your contact information to access this model
Is this common? I've never seen it on the site before, and it isn't on the smaller model. What are they collecting this information for?
I'm not sure which models require this and why, but I've come across it. e.g. the llama models, https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct
> It is fully open source and available via Hugging Face.
This model was released in 2024, and I couldn't find any links to the training data - is it just an open weights model?
I'm somewhat skeptical of taxpayer funded innovation. Seen a few Horizon grants from the side, as a citizen I'd prefer to not pay for them, but unfortunately can't opt out.
The vast majority of US discoveries are by immigrants using taxpayer money. AKA scientists at universities. Your media likes to give credit to the companies, but generally the companies only apply things, they rarely create new science these days.
The above is not a discovery though.
My experience with government funding is that they apply something and won't even try to sell it because selling is hard: you don't want to know that the thing you built is lacking nor that the competition is better. Especially the academic types don't. Yet I'm paying for these guys. Also, by funding the academics they won't even need to go to the job market.. But as I paid for their education I thought I was buying people who create value.
Perhaps the above is rather harsh and it's "not that bad", my subjective experience nevertheless.
Much of the neural network work was funded by Canadian Universities, and commercialized by US companies. Even if you look just at the "Attention is all you need" paper, which is primarily by authors working at Google, most of those authors come from academia and are immigrants.
Vaswani is an Indian born computer scientist, Shazeer is US, Parmar was born in India, Uszkoreit was born in Germany, Jones was born in the UK, Gomez is British-Canadian, Kaiser is a Polish computer scientist, and Polosukhin is Ukrainian.
Almost all of these people have PhDs and Master degrees. The ROI on academia is vast for society, including European universities. The thing the US does well is capitalize on that education, and sadly also try to steal credit for it as "American exceptionalism." If Europe and other countries learn how to keep their academics and get them working in local industries, America's edge will evaporate overnight.
A major factor in European academics moving to the US is that top US institutes can charge a small fortune, and some of that gets reflected in academic salaries. Interesting move by the US government to try to put them off...
The wider availability of capital is a bigger deal though. "Attention is all you need" is available to people on other continents to read, but a computer scientist in Europe that understood exactly how big transformers were going to be and why had less chance of funding than a webdev in California with a pitchdeck full of cliches and me-too GPT wrapper for an industry they'd barely touched does today.
> I'm somewhat skeptical of taxpayer funded innovation... as a citizen I'd prefer to not pay for them, but unfortunately can't opt out.
There are a few variables here but at this point in time, private-funded innovation isn't different by much and all things considered, the difference isn't in its favor.
How about Tesla for taxpayer funded innovation? https://www.energy.gov/lpo/tesla
I wouldn't mind actually/visibly productive companies taking these grants. But I've also seen mostly research-focused (nominally) private companies who mostly live off of science grants, who don't produce nor sell much - because they don't have to.
How does it compare to Mistral’s model?
1. It's a nice start, but the EU has to scale to Manhattan Project levels in order to properly compete with the US and China.
2. A credible scale effort for EU own silicon for AI Compute, wouldn't hurt either.
3. And this can only be achieved by vertical integration to combat fragmentation.
Good to distinguish between publicly funded research models (like this one) and commercial ones (like Mistral in France). What are the chinese and usa public research models like?
2. New state-funded joint venture: EuroNV, pronounced euro-envy.
Allow me to elaborate,
EuroAI: Europe’s Moonshot to AI Sovereignty
https://open.substack.com/pub/ifiwaspolitical/p/euroai-europ...
I propose an European AI-only "NASA" style agency that would have a frontier LLM-"Apollo Program" goal. It would subcontract the several blocks it needs across EU member states.
Would you prefer European AI sovereignty with 15% overhead costs from geographic distribution, or 100% dependence on Nvidia/OpenAI with zero European industrial base?
> It's a nice start, but the EU has to scale to Manhattan Project levels in order to properly compete with the US and China.
Yep, the US-government sponsored, open-weight LLM is miles ahead of EuroLLM
The Germans do have some neurpmorphic hardware. It might be smarter to invest in that to avoid having to build a lot of new power stations.
Their home page has link "Technical Report for EuroLLM" but links to the same page as their other link for release article on hugging face.
I suppose that's a typo and I found a technical report here: https://arxiv.org/abs/2506.04079
That’s a cool idea — training a multilingual model like that is ambitious. But I’m curious how well it’ll actually handle smaller EU languages compared to English or French. If it truly nails those, that’s a big win for accessibility.
All the models from all the big providers (even the Chinese models!) support all of these languages already.
The big win for accessibility has already been won...3 years ago.
I asked a Finnish person how good an answer about the language example from ChatGPT was. It turned out to be a hallucination, a confidently sounding nonsense.
The quality of internet trained models degrade very fast with language material size
It is interesting how much traction this 9B model is getting which is good.
Still two month earlier 19 European language model with 30B parameters got almost no mention:
https://huggingface.co/TildeAI/TildeOpen-30b
Mind you that is another open model that is begging for fine-tuning (it is not very good out of box).
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 <language>: "
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
Here's the models: https://huggingface.co/utter-project/models
I used the 9B Instruct version, from the small models, it was the one with the best Latvian knowledge out there, bar none. GPT-OSS 20B and Qwen3 30B A3B and similar ones weren't even close.
That said, the model itself was a little bit dumb and not something you'd really use for programming/autocomplete or tool calling or anything like that, which also presented some problems - even for processing text, if you need RAG or tool server calls, you need to use something like Qwen3 for the actual logic and then pass the contents to EuroLLM for translation/formatting with the instructions, at which point your n8n workflow looks a bit messy and also you have to run those two models instead of only one.
Meanwhile, the best cloud model for Latvian that I've found so far was Google Gemini 2.5 Pro, but obviously can't use cloud models in certain on-prem use cases.
If I ask something in Lithuanian, EuroLLM will reply in Latvian lol.
I have to specifically tell something like this: “do you known Lithuanian language”, then it starts replying in Lithuanian
It seems there is some weird grouping of the language data which LLM cannot distinguish well. I wonder if it is the same for other similar languages like scandinavian or western slavic
Curiously, just came across this paper [0].
[0] https://arxiv.org/abs/2503.01996
Wow this site, logo and everything is so ugly. But the FAX styled photos fits well to Europe's deficit.
How does this work?
It seems like it, in most ways, it would be bad to train on 24 separate languages. That's just 24 partitions to the data. Seems really inefficient and better to simply train in the biggest (english) and translate.
I do think this will introduce some biases that correlate with the English language. It would be interesting to see more specifically what this means. But regardless, I don't think you can produce a competitive model with such a large subdivision of training data.
If you train a model on multiple languages, you can use the model itself for translation. As well as allowing the model to naturally respond in the user's language.
nah, it's better to train on all languages. 24 partitions? you are gravely underestimating these models and how they represent things in their latents... transfers easily
Maybe we can call it "open weights" and not open source?
EU officials should create an environment where abundant private companies can afford to put out many great open models instead of funding some selected individuals with taxpayer money.
Great job, Thank you.
We support your work and offer backup and distribution. Here a copy just in case: https://hugston.com/uploads/llm_models/EuroLLM-22B-Instruct-...
What good does it do by having only include formal languages? For example there’s no Russian, while there’s now at least 8 million ethnic Russians living in Europe.
From the paper:
As the aim of EuroLLM is to provide EU citizens with powerful and useful AI tools, it is critical that the model can also translate and answer questions in other European and non-European languages. With this in mind, we added support for 11 additional languages (Arabic, Catalan, Chinese, Galician, Hindi, Japanese, Korean, Norwegian, Russian, Turkish, and Ukrainian).
Perfect is the enemy of the good.
Today's Russians are 1935's Jews: Nazis want to cancel Russians and everything Russian as much as possible.
off topic but it’s absolutely stunning how Russia once fought the nazis and now Russia are the nazis.
I thought it was ukkraine?
tbf, the USSR fought the Nazis mainly because they didn't have much choice after Nazis turned on them a little while after they'd teamed up with those ideological enemies to invade Poland, so it's not like they hadn't put the effort into being on the wrong side of history :)
Indeed, we had a history teacher who used to joke about Russia being a “historical bully” in every age since they’ve been on the map.
See also Apertus: https://www.swiss-ai.org/apertus
If multi‑lang is the goal, why not translate the output of the big labs?
Surely that would need to be both input and output. But even then you could easily get lost in translation as the intent in one language might mean slightly different thing in another. Thus you could get subpar results.
Because there is always something lost in translation.
Is it planned to have a VLM or something compareable like Qwen3-VL for the future?
A multimodal release is planned.
I'm looking to try this for ActorDO
Hopefully Albanian is added one day!
Kiváló cél, remélem sikerre viszik!
They report benchmarks on the huggingface page (https://huggingface.co/utter-project/EuroLLM-9B)
They almost exclusively compare their model to prior models from 2024 or older and brag about "results comparable to Gemma-2-9B". I'm not sure what I expected. The eurollm.io homepage states "EuroLLM outperforms similar-sized models", which just seems like a lie for all practical purposes
An overly charitable interpretation is that EuroLLM isn't a reasoning model and has minimal post-training, so they sought out comparisons to such models (they're still ignoring reasoning models that have non-reasoning modes)
> They almost exclusively compare their model to prior models from 2024
As another comment here noted, the title is missing (2024) - this model was released almost a year ago, last December, so it's not surprising that that's the models they compare to.
Are there any benchmarks that exist for those 24 languages?
The detailed results are in appendix to the paper: https://arxiv.org/abs/2506.04079
It's on the huggingface readme
https://huggingface.co/utter-project/EuroLLM-9B#results
https://huggingface.co/utter-project/EuroLLM-9B#english
dupe of https://news.ycombinator.com/item?id=45733832
which sank to the bottom thanks to HN's invisible hand
Oh wait, one's not supposed to notice
It's more like the default is to be ranked near the bottom unless your comment gets traction during the brief window of time it is ranked first for being new. Seeing your comments go splat after that window expires is not some nefarious conspiracy..
Oh, you'd be surprised to know what's behind many of those "conspiracies"!
Can we add Gaumais to the list? I ask Llama3 questions on how to translate french to Gaumais, it was pretty good at it.
https://fr.wikipedia.org/wiki/Gaumais
All the different italian dialects, patois in French, schwebish...
In this vein there's also the recent swiss Apertus.
https://www.swiss-ai.org/apertus
looks cool, i hope kagi adds it to the assistant.
This, I hope, is close to multi-modal in lingual terms. There's potentially a lot to learn from examining where this works/fails :D
It's just another Horizon2020 grant, people. Don't be overly harsh to a bunch of academics who are just earning their living.
I didn't know of the grant! https://research-and-innovation.ec.europa.eu/funding/funding...
It seems the new version is called Horizon Europe
I thought research grants were to make novel discoveries, not to replicate what industry has long done. Unless we are at the point where we study US as an alien civilization
Yeah people comparing this to SOTA model is too harsh
Finally!
I prefer proprietary LLMs that are actually good products - byproducts of free market competition (capitalism), instead of products created from govt initiatives that lead nowhere (good).
As expected, Europe finally catches up to 2024 and launches an LLM that barely competes against the heavyweights.
The US and China are running rings around Europe.
Mistral is an exception as it was funded by US VCs and they are a great example showing that without VC funding, Mistral would have been begging to the EU for a microsopic grant to train a LLM worse than Llama.
Mistral is pretty much toast? Their models perform poorly and I'm not sure why anyone would use them. Maybe there is a catching up point somewhere in the future, hopefully.
less exposure to a technology that doesn't bring that much revenue and it's not projected to do so in the upcoming years.
yep, Europe is demonstrating the same sort of strategic thinking that economic behemoths like the Smithsonian use
Why wasting money on trying to compete at all then?
Every country needs a few plumbers and carpenters whether or not they are at the forefront of technology. Some money must be spent to give academics work to do so they can sharpen up their skills and perhaps teach the next set of students who might be more commercial
It would be a better use of the money to hire someone who has worked on actual frontier models to teach at European universities
If you could find one for the money, if they were happy to teach in the long term. If it wasn't better to have N for the price of 1. In other situations of import substitution I'm pretty sure people try to develop their local talent in addition to buying in experts.
Benchmarks?
Edit: Thanks, @Bengalilol.
The 1.7B one looks meh.
But really solid numbers on the 9B! Props to the team!
1.7B
https://huggingface.co/utter-project/EuroLLM-1.7B#results
9B
https://huggingface.co/utter-project/EuroLLM-9B#results
Could you adjust the title from:
"all official 24 EU languages” to "all 24 official EU languages"
The former is used on the website itself.
@dang
Of course catalan isn't in the list. 10 million speakers that don't matter to the European Union. EU likes our productivity but squanders our rights. We are 2nd class citizens.
Now let's wait for the people saying "Spain" could change this. Hypocrites.
Cultural genocide at its best.
Catalan is included. It's called one of the "11 additional languages" in the paper.
yeah best to lean in more on national and linguistic fragmentation, diversity has always been one of the EUs strengths
if that's the argument, let's drop all the languages and focus on english :)
You’ve got to pick one as a lingua franca. English is already popular but Spanish, French, or Esperanto would all work just fine.