47
Here is a much better way for Europe's tech firms to catch up in global AI race (spoiler: a multilingual, fully open source, law-compliant, democratic and homegrown LLM): https://slrpnk.net/post/17978607
this post was submitted on 03 Feb 2025
47 points (87.3% liked)
Europe
2004 readers
624 users here now
News and information from Europe πͺπΊ
(Current banner: La Mancha, Spain. Feel free to post submissions for banner images.)
Rules (2024-08-30)
- This is an English-language community. Comments should be in English. Posts can link to non-English news sources when providing a full-text translation in the post description. Automated translations are fine, as long as they don't overly distort the content.
- No links to misinformation or commercial advertising. When you post outdated/historic articles, add the year of publication to the post title. Infographics must include a source and a year of creation; if possible, also provide a link to the source.
- Be kind to each other, and argue in good faith. Don't post direct insults nor disrespectful and condescending comments. Don't troll nor incite hatred. Don't look for novel argumentation strategies at Wikipedia's List of fallacies.
- No bigotry, sexism, racism, antisemitism, dehumanization of minorities, or glorification of National Socialism.
- Be the signal, not the noise: Strive to post insightful comments. Add "/s" when you're being sarcastic (and don't use it to break rule no. 3).
- If you link to paywalled information, please provide also a link to a freely available archived version. Alternatively, try to find a different source.
- Light-hearted content, memes, and posts about your European everyday belong in [email protected]. (They're cool, you should subscribe there too!)
- Don't evade bans. If we notice ban evasion, that will result in a permanent ban for all the accounts we can associate with you.
- No posts linking to speculative reporting about ongoing events with unclear backgrounds. Please wait at least 12 hours. (E.g., do not post breathless reporting on an ongoing terror attack.)
(This list may get expanded when necessary.)
We will use some leeway to decide whether to remove a comment.
If need be, there are also bans: 3 days for lighter offenses, 14 days for bigger offenses, and permanent bans for people who don't show any willingness to participate productively. If we think the ban reason is obvious, we may not specifically write to you.
If you want to protest a removal or ban, feel free to write privately to the mods: @[email protected], @[email protected], or @[email protected].
founded 7 months ago
MODERATORS
Deepseek is welcome in Europe as all others, as long as it complies with EU's GDPR and the law: A quick reminder that Deepseek is being probed so far in Italy (where it's prohibited), in France, and Ireland. We'll see whether other countries follow.
It's not the DeepSeek service that's providing a huge opportunity, it's the model. That can be run locally without any sort of privacy concerns.
Sure. But it's a black box, no? That model might as well have been trained to sabotage in subtle ways after several weeks.
If developers are asking for code help and copy-paste without really understanding all the pieces, it seems like it could get real bad.
Edit: Am I technically wrong, or do you think I'm just overblowing the risk?
The model has no idea how much time has passed unless it is explicitly told what time has passed. They're not capable of forming new memories during routine operation, the black box remains immutable unless you explicitly do additional training on it - in which case you're supply it with the training materials yourself and you know exactly what's in them. People who use LLMs for coding already know they're not perfect, and they're not going to be all that helpful unless you know enough of the programming language to know what it's trying to do. I don't think the sort of subtlety you're suggesting is really possible to train into an LLM with our current technology level.
And even though it's a black box, it's not magic. It can't communicate with the outside world in any way other than the ways you provide it, and it can't do anything unless you're actively empowering it to do something.
So I'm not really concerned that DeepSeek has some kind of super secret hidden "programming" that's going to jump out and stab us. I think its only "threat" is what we already see on the surface - it's hugely disruptive to the business plans of companies like OpenAI, who were betting on AI remaining a hugely expensive and centralized affair.
Can we like stop copying shit from America?
We don't need European tech giants on global scale, that never works out well for the people.
Also the slow rise of fascism all over the Europe makes me far more worried than the AI race (which is silly, AGI is still far away).
Yes we do. We're now in a situation where schools, companies and government agencies are using Google or Microsoft to host their entire organization, and Amazon to host their services. What happens when Trump decides to introduce digital tariffs on these products? Or declare war on Denmark?
I think you misunderstand the point. Of course we need more technical capabilities in Europe. But we don't need the companies providing said capabilities to have the influence of Meta, Alphabet and co. And influence comes with size. Replacing a US based monopolist with one headquartered in the EU doesn't get rid of the monopoly.
So no, we don't need an European tech-giant. What we need is more in-house know-how and more medium sized companies.
Just don't use them?
It's not like there aren't other options. In fact, it was Microsoft which used money to stop alternatives,
I'm not in a position to make these decisions on behalf of all companies, schools and government agencies that use their services in my country.
Then it must be equally logical that european tech would not be able to, either.
So what exactly are you trying to say? What was the initial argument?
Uh... You kinda do if you don't want your politics to be controlled by American broligarchs.
Block X, Facebook, and never look back.
It's not like there aren't alternatives, or like we need them at all.
It's not like there aren't alternatives
Things like DeepSeek are exactly the "alternatives" you're talking about here. DeepSeek provides alternatives to American AIs,
or like we need them at all.
People use what they want to use, and lots of people want to use social media. You're on a social media platform right now as we discuss this.
You will not catch me defending LLMs anywhere, on fact I'm dramatically opposed to them. Your argument does not touch the need to create european tech giant.
"You use social media, and you hate social media??!"
Yes, I hate centralized social media. As there is no need to use Facebook, TikTok etc
In related news:
Using algorithmic jailbreaking techniques, our team applied an automated attack methodology on DeepSeek R1 which tested it against 50 random prompts from the HarmBench dataset. These covered six categories of harmful behaviors including cybercrime, misinformation, illegal activities, and general harm.
The results were alarming: DeepSeek R1 exhibited a 100% attack success rate, meaning it failed to block a single harmful prompt. This contrasts starkly with other leading models, which demonstrated at least partial resistance.
CNBC reports that DeepSeekβs privacy policy βisnβt worth the paper it is written on.β
Seems to be a long way to go, but Hugging Face developers are in the process of building a fully open reproduction of DeepSeek-R1 as the AI is not Open Source as it claims.
So, is censorship a bad thing or not? This "safety" test is really just a censorship test and I consider "failing" it to be a good thing. I loathe when a computer refuses a command I give it because it thinks my command was "immoral".
I'd say we need uncensored models. Eric Hartford wrote a long blog post about this: https://erichartford.com/uncensored-models
And I'd have to agree. It's probably unhealthy to have some disruptive technology solely in the hands of some big companies who then get to decide how to shape the world with it. That's deeply undemocratic. And comes with lots of severe issues. We kind of need a more level playing field and a say, if we don't want to just be manipulated by the technology. But read the article, my few sentences here aren't as good.
In addition to my comments, we can add that Wiz Research uncovered exposed DeepSeek database leaking sensitive information, including chat history.
TLDR: DeepSeek had left over a million lines of sensitive data exposed on the open internet, including digital software keys.
That's DeepSeek the service, run by the Chinese company out of China and subject to Chinese jurisdiction. Not DeepSeek the model, which is what European companies would be making use of to catch up.
Oh no, models will be more responsive to anyone as opposed to only billionaires.
This is not good news, but when you've let the genie out of the bottle, this just seems like balancing the scales. At this point, transparency, not closing off the information to a select information, is a good thing. Something social networks like this fail to get.