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

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

eurollm.io

Post reply on HN
Print view View original

371–380 of 629 posts

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

#371

Post by IshKebab »

troupo wrote: post #314
IshKebab wrote: post #268

Earlier quoted context omitted.

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,…

Yeah you just demonstrated how it was misleading - by omitting half the categories, especially communications and culture.

> 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.

Riiight.

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

#372

Post by saubeidl »

cbeach wrote: post #351
saubeidl wrote: post #296

Earlier quoted context omitted.

[flagged]

The EU Commission is appointed, not elected. Only the Parliament (MEPs) are elected. What’s worse, the parliament cannot originate law. Only the unelected Commission can do so. And they can do it behind closed doors. This is a setup that’s ripe for corruption.

The Commission is appointed by elected officials. That's the same way the US presidency works. It's also how the UK PM role works, or any minister in pretty much any democratic government. All of those are still referred to as "elected" in common tongue.

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

#373

Post by troupo »

vanviegen wrote: post #365
troupo wrote: post #346

Earlier quoted context omitted.

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...

What does it matter that the rules for establishing differ per country? I'm only founding in one of them. The article is unclear, but is probably referring to making it easier for startups to offer products in other EU countries.

The idea is to establish common rules to make it easier to register and move startups between countries, among other things.

It's in very early stages, so info is very scattered. More info, for example, here: https://www.loyensloeff.com/insights/news--events/news/the-2...

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

#374

Post by rgblambda »

lawlessone wrote: post #243
RobotToaster wrote: post #97

Earlier quoted context omitted.

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?

Anglo-Saxons not Dutch. But the short answer is yes. The word Welsh is derived from the Old English word for foreigner.

Latin would have been spoken in towns and cities but as Roman rule collapsed it was replaced by Brittonic (ancestor of Welsh), unlike in the continent where it developed into various Latin derived Romance languages.

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

#376

Post by jacobgorm »

sealeck wrote: post #328
carlosjobim wrote: post #313

Earlier quoted context omitted.

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...

Denmark has a large hearing aids industry due to lots of government funding for hearing aids, and a large wind turbine industry due to funding for wind farms. So stimulating demand can work to build or strengthen an industry, but what Denmark and EU are doing with GPUs is stimulating supply in Europe and demand in the US. I would be surprised if that does not end up strengthening US and not EU industry.

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

#377

Post by airspresso »

loandbehold wrote: post #299
tensor wrote: post #112

Earlier quoted context omitted.

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 keep hearing that LLMs are trained on "Internet crap" but is it true?

Karpathy repeated this in a recent interview [0], that if you'd look at random samples in the pretraining set you'd mostly see a lot of garbage text. And that it's very surprising it works at all.

The labs have focused a lot more on finetuning (posttraining) and RL lately, and from my understanding that's where all the desirable properties of an LLM are trained into it. Pretraining just teaches the LLM the semantic relations it needs as the foundation for finetuning to work.

[0]: https://www.dwarkesh.com/p/andrej-karpathy

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

#378

Post by jacobgorm »

jll29 wrote: post #349
dzikimarian wrote: post #234

Earlier quoted context omitted.

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 obt…

Someone should build a startup that uses the EuroLLM to generate EU funding proposals.

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

#379

Post by rat87 »

adzm wrote: post #2

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 ar…

Why Italic as opposed to Romantic/Latin? I don't think there are any surviving not Latin branches of the Italic family are there?

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

#380

Post by airspresso »

isodev wrote: post #290
sajithdilshan wrote: post #138

Earlier quoted context omitted.

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…

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.

I agree. And I also know how much of that experience comes from having a legal dept. that are collaborative and supportive of what the tech org wants to do. Which I suspect is quite rare.
Post reply on HN