alt Hacker NewsCan Europe train a frontier AI model on the compute it owns?
Comments
The question was never do enough computers exist in Europe, but rather can Europe organize the capital and cross-company / cross-border relationships required to build a big model at scale. There the answer still looks iffy at best. This is where the US has, and continues to, thrive and where Europe can’t get out of its own way.
Can Europe train a frontier AI model?
It seems unlikely at the moment. The writeup you have is quite interesting, but notably distributed compute for frontier is at least a bump in the road.
As other's have pointed out, if resources across Europe came together there could be a good amount of resources today to do so, but it is fragmented.
Project Glasswing, and most recently the blocking of Fable 5 potentially only for non-US citizens have sparked more attention in the political spheres regarding sovereignty/competitiveness of European AI.
There as also this great write-up which explains the trajectory Europe currently has when it comes to AI compute here: https://europe2031.ai/
So theoretically: yes, but there doesn't seem to be a big enough will to attempt to catch up with the trajectory of American hyper-scalers.
Sure, theoretically, if it could come to an agreement, and meanwhile the cross-border cooperation of even the inter-EU countries is at an abysmal rate, and currently, even during a goddamn land war with Russia, Germany and France can't come to an agreement to build a fighter jet ensemble (together).
What is this whole thing about Europe being behind on AI? Do Mistral and DeepL not exist? Yes, I know DeepL is niche, but IMHO it is the best translation model out there.
Forget regulations, would it make sense for Europe to train a frontier model of it's own? Would it be sufficiently better than fine tuning a Chinese model? Would it actually be competitive with US frontier models? Would enough people pay to use it even within Europe to pay for the training costs? Do we have enough inference capacity that enough people /could/ use it? Would being "European" allow any governments in Europe to trust it, rather than deciding that actually there needs to be a French, German, Italian, Spanish and UK sovereign AI?
I am guessing that enough of these questions can be answered with "no" that nobody really wants to invest.
For the same reason there isn't really a serious third start up competitor to OpenAI and Anthropic.
Why even bother creating a repo like this? Why not just link to a ChatGPT conversation?
Under normal circumstances, I'd agree with most folks here, it'd be highly unlikely.
However, we're (i think officially) in an arms race.
I wouldn't want to bet against anyone in these unprecedented times (with plenty of historical parallels).
Being slow is sometimes a problem.
But if the needed total revenue for an investment is so absurdly large, that we need to capture 5-10% of the labour market, being slow might mean not burning in the crash.
The US drove the CCCP to bankruptcy by investing in nuclear bombs. Maybe history repeats itself.
Does Europe need to train a frontier AI model, at all?
"Training a frontier AI model" is a euphemism. The AI industry is, at present, a disproportionately resource-intensive and exploitative activity which carries only an religious promise of making up for itself at some future date (note that the beliefs being touted to board members are different than the gospel spread to users).
More accessible "non-frontier" models are being designed, built, and trained, while bigger models' gains are beginning to asymptote. Economically, Europe does not often hastily participate in new ways of wealth extraction, and with regard to sovereignty, I'd argue that "frontier" models are becoming a liability to the governments who build them and to the populations they're meant to lead.
No because public EU compute is adminisered by physicists and although they are very competent in their own domains they are winging it when it comes to AI. It like expecting OakRidge to train LLMs...
They already tried training LLMs.
I feel like the very thing that EU is great at: more consumer friendly and anti-corporate legistation, also hamstrings it for innovation. Why would VCs invest in AI there instead of in the US where they don't have to worry about any of that.
This has factored out product development, which is more than compute resources. Just like any industry, some organisation needs to take ownership and responsibility to convert technology to a usable product.
Just post the prompt that generated this slop next time. Then we'd have a chance of seeing some original thoughts, instead of a bunch of web searches filtered through a bucket of mediocrity.
They could, but why? US and China has poured Trillion of USD into training and any semblance of getting those money back seems like some far fetched dream. Currently there is no realistic path to profitability with these models.
What I see as usable product in the future are smaller specialized models which are able to run and be trained with fraction of resources what goes to current frontier models.
> Can Europe train a frontier AI model on the compute it owns?
No
Are we really discussing a plan that has never been attempted before (training a frontier model on federated hardware) and that would require coordination on a continental scale, sketched on a five-page PDF on GitHub with no discernible author or affiliation? a PDF that, I am sorry to point out, reeks of AI prose in more than one passage?
What are we even doing here.
Who cares? Just distill some existing frontier models and run the inference yourselves. Instant sovereign AI.
europe2031.ai
They really shouldn’t.
those that know, do not talk?
[dead]
[flagged]
[dead]
No, it cannot. The fundamental problem between the US and EU it is that in the US if something is not explicitly forbidden, it is allowed and in the EU if it is not explicitly allowed it is forbidden.
You cannot have innovation at the speed and scale that you have in the US because the legislation is cumbersome and there is no unified market with the same rules.
Absolutely not, even if Europe was given all of the compute in the world. The issues are much much worse. Starting at "data privacy", continuing with the "EU AI Act" and just an overall mindset of regulation and German angst. Many structures in the EU actively prevent and fight against innovation. Oftentimes more subtle but the consequences stay the same. If you ever had to use one of the "sovereign" AI provides such as StackIT, OVH and the-like, I feel with you. It is just so bad in terms of product and performance, there is no comparison at all with Hyperscalers, and it shows. Eventually, it's a cultural and structural problem along the way and the future here looks horrible.