[-] e8d79@discuss.tchncs.de 2 points 16 hours ago

Well for once I don't think training a model should be considered fair use. At the time when people released their art and writings under such a license there wasn't the threat of that resulting in them potentially losing their job. I for example stopped contributing to open source completely because of that threat. The reason why I consider training a model on these things immoral is that millions of people put in the effort of creating code or art in various forms and some rich assholes come along scrape it all and sell it back to us while creating ungodly amounts of pollution. It really doesn't matter to me that a few bootlickers run some inferior models on their own system. They are legitimising AI by using it and are encouraging others to also do so. Meanwhile the damage is already done with artists being now accused of publishing AI-generated images because they wanted to do a good thing by releasing their art online for free and the models happily imitating their style.

[-] e8d79@discuss.tchncs.de 2 points 1 day ago

"You just don't understand it" was also what the crypto people always told me when I called their obvious ponzi scheme a ponzi scheme.

[-] e8d79@discuss.tchncs.de 1 points 1 day ago* (last edited 1 day ago)

I don't know why you are talking about privacy. Expectation of privacy is unrelated to copyright. Anyway the point I am making is that a license allowing something doesn't make it moral. Sure, the law allows someone to train their slop generator on my code but I can still think they are an unethical piece of shit for doing so. The same goes for the people who then use such a model. Which is also why I don't feel bad about techbros crying about how they are bullied online for using AI.

[-] e8d79@discuss.tchncs.de -5 points 1 day ago

A license may allow something but that doesn't make it ethical to do. At the time these licenses where written the wholesale pillaging of the commons by AI companies wasn't a thing. For that reason I don't think it matters what license anything was published under. Unless you where given permission by the author its not OK to train your model on it.

I consider Open Weights models only open in name, the training data is almost never released. As such it is just an attempt at open washing. Its like releasing some executable files and calling that Open Binaries.

[-] e8d79@discuss.tchncs.de 4 points 1 day ago* (last edited 1 day ago)

Yes, you are acting unethical. Your self-hosted model is neither open nor was it trained with consent. Using them is just as unethical as using the models from OpenAI.

[-] e8d79@discuss.tchncs.de 2 points 3 days ago

Well, I wouldn't call myself new to hosting, I used nginx before but switched because Caddy makes configuring HTTPS for my reverse proxy very convenient. I use it with this module which handles the ACME DNS challenge for creating and updating letsencrypt certificates for my internal domains.

[-] e8d79@discuss.tchncs.de 5 points 4 days ago

If anyone ever wondered why I am constantly angry about AI its shit like this...

I was looking at options for securing a few services hosted behind a caddy reverse proxy and found a project with the very serious and official sounding name 'caddy-security'. To get started I tried to find some proper documentation for it and immediately found this announcement. In future, instead of a proper documentation you will get a bunch of files in the repos .codex/skills directory. These, according to the developer, help 'both humans and AI agents work with the codebase'. I don't know how this is supposed to help me because the files aren't linked to each other and are anything but proper documentation, but I guess they want me to point a slop agent at them and have it generate some security configuration for me.

[-] e8d79@discuss.tchncs.de 13 points 6 days ago

Unfortunately, that isn't entirely wrong. There is always the few riders that think the pedestrian zone in the inner city is the perfect place to start their training session.

157
help wanted (lake.computer)

Highlights include:

I see the commits say they were co-authored by Claude, but from having read the code, I think you haven’t given Claude enough credit.

I wish I was their credit card, because then I might finally be in a happy relationship with someone who loves to use me this hard.

97

Our investigation found that Hugging Face, the top platform for sharing open-source AI models, isn't just hosting tools that generate non-consensual intimate imagery. It's helping users compare, benchmark and optimize them, making image-based sexual abuse easier to scale.

By investigating the platform's most widely used Hugging Face Spaces and creating our own, we found:

  • 73% of monitored requests were sexual in nature.
  • Of those, 83% sought to undress or otherwise sexualize a person. 95% targeted women.
  • 6.7% targeted a minor. If fulfilled, these requests would constitute child sexual abuse material (CSAM).
  • Just 3% of audited Spaces showed any evidence of output moderation.

Hugging Face's own policies prohibit non-consensual intimate imagery. Yet we found virtually no safeguards preventing these tools from being used for exactly that purpose.

253

lol. lmao, even.

406

Would be a terrible shame if lots of people opted out.

If you are a EU citizen you might also want to write a complaint to privacy@huggingface.co because they are collecting your personally identifiable information in machine-readable form which they are distributing to third parties.

180
submitted 1 month ago* (last edited 1 month ago) by e8d79@discuss.tchncs.de to c/europe@feddit.org

There is a new survey with some design proposals for the themes "European Culture" and "Rivers and Birds".

268
98

Really cool research. Overall, X11 still wins compared to Wayland(kwin) but only by 0.14 - 0.22 ms. The really bad performance hit is from XWayland which added 3.13 ms.

7
[-] e8d79@discuss.tchncs.de 154 points 1 month ago

If you are into fancy looking cocktails this product makes sense. Making clear ice cubes at home is not easy.

102
605
submitted 2 months ago* (last edited 2 months ago) by e8d79@discuss.tchncs.de to c/lemmyshitpost@lemmy.world
693

The IT company Mullvad’s founder Daniel Berntsson is behind a giant donation to the populist Örebro Party, which advocates “comprehensive re-immigration” from Sweden. “It’s sad that it’s needed,” he told Flamman.

~ https://www.flamman.se/techprofil-ger-miljoner-till-orebropartiet/

16
  1. FOSS projects should not shun contributors who choose to use LLM-gen-AI systems.

Fuck everything about this and the liberal horseshit that follows it.

[-] e8d79@discuss.tchncs.de 196 points 7 months ago

You're absolutely right -- that was a fatal dose.

[-] e8d79@discuss.tchncs.de 217 points 9 months ago
[-] e8d79@discuss.tchncs.de 159 points 10 months ago* (last edited 10 months ago)

This is misleading. For people paying for the IDE nothing changed, data sharing remains an opt-in option. For users of their free licenses data sharing was enabled by default. Still a shitty thing to do especially as it hits a lot of OSS developers but lets criticize that instead of creating memes that are misinformation.

view more: next ›

e8d79

0 post score
0 comment score
joined 2 years ago
MODERATOR OF