Interesting idea, but there's already some really good Fediverse clients (such as Fedilab). Adding AI translation is unique, but it rubs me the wrong way that it would be potentially submitting people's toots and comments to a LLM without their consent.
this post was submitted on 01 Aug 2024
66 points (98.5% liked)
Open Source
30768 readers
556 users here now
All about open source! Feel free to ask questions, and share news, and interesting stuff!
Useful Links
- Open Source Initiative
- Free Software Foundation
- Electronic Frontier Foundation
- Software Freedom Conservancy
- It's FOSS
- Android FOSS Apps Megathread
Rules
- Posts must be relevant to the open source ideology
- No NSFW content
- No hate speech, bigotry, etc
Related Communities
Community icon from opensource.org, but we are not affiliated with them.
founded 5 years ago
MODERATORS
Would you feel the same way if it was being sent through a "non AI" translation server?
Yes, because the core problem is that taking somebody else's entire comment and putting it into a service neither I nor the dev controls is to violate the original author's consent. The original author has no way to provide consent. Likewise, if I used a live translation service to communicate with somebody in real time, I would want to somehow first verify they were okay with me using it.
"If a product is free that should otherwise cost money, you're the product."
One could argue simply posting something online is consent, especially when in a publicly accessible location.
If you post something on Twitter, and someone else screenshots and posts it elsewhere, are they "violating consent" what if that person was Donald Trump or some other ghoul saying some crazy shit. Would you have the same reservations?
This seems like an extreme take to me on what communication and consent mean on the internet.
One could argue simply posting something online is consent, especially when in a publicly accessible location.
That's exactly the argument of the companies slurping up online data. The problem is that not explicitly revoking consent ≠ granting consent. It's the same argument employed by rapists. "They didn't say no..." and obviously, we recognize that extreme example as fallacious reasoning (specifically Denying the Antecedent).
- Let C be "denial of consent."
- Let L be "use by LLMs."
- C => !L ✅
- !C => L ❌
If I post something online, I'm not defacto granting that I want a machine or a corporation using those words for their gain, and that likewise applies to anyone who does not expressly grant consent to use their online interactions for someone else's profit.
Are we really comparing commenting online to rape now? That's a huge leap
These are public sites that are used for free I don't think there's really any expectation of privacy, additional translation software is far from a nefarious thing.
You think the Fediverse is free? You do know it takes labor to develop, right? You know servers aren't free, right?
And anyway, privacy isn't the issue. Consent is. Public writings are by default not private.
But since you ignored the very prescient example of the argument used by rapists, here's another one that shouldn't make you balk and dismiss out of hand:
First, suspend the legality of this analogy. We're discussing logic, not law. Next, pretend I take your phone that you left on a coffee shop table. You say, "That's wrong. Give it back." I say, "You didn't expressly say I can't take it, and it's sitting there in public, therefore I can take it." It doesn't follow that I can take it just because you didn't explicitly shout out to the other patrons, "Don't take my phone!"
It's the same with anything you put out in a public forum. Leaving it there isn't implicit consent to use for someone else's gain.
It uses OpenAI for translation. Spyware (unless it uses the DDG implementation or something).
How would the DDG implementation be less than a privacy concern? It's just adding a middle layer. It may only lump queries together in a larger pool of seemingly undistinguishable users. Am I missing anything else?
They made a deal with the AI providers to not use the data for model training and it has some open-source models now too (though idk what providers do they use).
Spyware because of a feature you don't even need to use? Geez.
It violates privacy and sometimes copyright of the user whose content you're trying to translate. And you can just press it accidentally.
Yes.
"Sure this option installs malware on your machine, but you don't need to use it. Geez"
I love that it's AGPL licensed!
Also, the associated website could use some attention when viewed with Firefox on Android.
...use of AI tools to...
yeah, no thx.