You must log in or # to comment.
The “race” to lose as much money as possible when the bubble bursts.
“The only winning move is not to play” - WarGames
It’s yet another thing about artificial intelligence that science fiction tried to warn us about. Perhaps not quite in the way it intended but it‘s prophetic nonetheless.
Should they? I mean I guess it’s better to use your own models and servers than add yet another dependency to the USA, but there are SO many issues with AI.
Yes, they should. I guess this applies to most technologies. Where is a coal plant/nuclear plant/data center/chemical plant/battery plant etc. most sustainable? Probably in the EU rather than the US or China or India.
There are several excellent LLM projects in Europe.
The EU released recently released a multilingual AI model for all 24 official languages in the bloc.
Italy came up with Minerva, the first LLM that was trained on data in native Italian.
There is also Ellis, a cooperation network.
So, I guess Europe can and will, and it also has to hold its own AI in my opinion.
Europe is certainly behind the Frontier Models from OpenAI, Anthropic, and Google. China is close behind and we should be happy that they publish their models. Mistral does not seem to have the ambition/budget to challenge the state of the art.
We do have some nice specialist model in Europe. We guess we have the data centers although they are unfortunately owned by US companies mostly.
So, I’d say no we can not hold our own. We are second class. That is not catastrophic but also nothing to be proud of.
Europe produces excellent research and is a world leader in many scientific areas, including in technology, but has historically struggled to scale startups. The former is definitely something Europe can be proud of, the latter has to change.
Regarding LLMs, the EU may have a key strength for future development if and when it maintains its regulations regarding data protection and privacy. I thing that neither the U.S. nor China will be a trusted source in the foreseeable future.
As the French Nobel economist Philippe Aghion says, "I believe there will be a demand for a more ethical AI, an AI that avoids some dangers … “People want to be protected. To have some regulations will make us [Europeans] even more attractive.”
Edit: I found an interview from late last year with Olivier Debeugny, the CEO of French start-up Dragon LLM.
In brief, he argues Europe should build its own AI architecture, not rent American capability indefinitely. He draws an analogy to Airbus: Airbus succeeded because European governments decided sovereign aerospace capability was worth building collectively, Debeugny argues. Europe should apply the same logic to generative AI. The argument is that dependency on foreign labs is a structural liability, not just a cost line.
Isn’t Soofi the best open source and best fully open source model at the moment? It’s a German European model.
Soofi is not at the frontier with the closed models. But given the low budget it was built with compared to the closed models, I don’t think that Europe must remain second class for very long.
It is of course in the interest of the incumbent firms to convince the world that this is a race and that others can never catch up or overtake them.
I don’t think that Europe must remain second class for very long.
Yes, Sofi is a great project. They may not yet catch up, but it isn’t their last release. Europe has all what it takes to catch up in all areas, and LLM is one of them.
TL;DR
nobody knows
It’s. Not. A. Race.
Of course at some point it becomes a self-fulfilling prophecy - a bubble. And not all bubbles burst, unfortunately.
