hendrik

Always a breeze once I go to a music festival or bigger event and there’s all the diversity with the German language. I think there are many places like that. And even in the larger metropolitan areas you can tell the difference between Cologne, Düsseldorf and the Ruhrgebiet and the people slightly to the east or north of it. At least where they grew up because all of it mixes in the cities and people will also commute 1h to work. So I think it happens in villages, cities and everywhere. It’s not entirely the same, though. Seems to be more nuance here than proper dialect, but people from 3 cities away will occasionally tell me on how my grammar has some funny peculiarities.

What’s the encryption and signing on a hardware level for? I mean dependent on what’s that good for and who controls it, it’s trusted computing, or treacherous computing as Stallman calls it…

(I mean it’s not working out great for GrapheneOS either. Back in the day I had a phone I owned, with privacy features added and alternative background services so I had a pretty much Google-free experience. These days it’s all locked down, I hand out my private metadata to Google, can barely ride a train without, or get a discount in the supermarket. I can’t do backups and I’m f***ed if I want to cross a border to a more restrictive country because these guys are in on it as well. They’re probably going to use it to limit what I can install. And more and more manufacturers lock down bootloaders etc and I thought we were past this. Graphene itself advised me to switch to proprietary code in the name of security and they’ll have a look at the code later, once Google eventually releases it. All of this is due to (or related to) these security measures working way too well and that’s also why they’re being used. I wish my phone didn’t have a TPM but a simple disk encryption like LUKS on Linux instead. And I don’t see many reasons why we should copy these very bad dynamics.)

I think the overall idea is nice, though. We had these project ideas to just plug in a box and be self sufficient in the self-hosting community since the SheevaPlug. Or the FreedomBox. There are some hardware projects as well like the Home Assistant Green or back in 2019 they tried to sell a Pioneer-FreedomBox. None of those match exactly with your proposal, but I think they’re pretty close. Maybe get in touch with them and see if you can participate in a new iteration, or read about their past experience with the proposed target audience. Especially FreedomBox seems like a good fit to me. They’re not very loud, but afaik still around. And they’re Free Software nerds, which seems to align with your idea, minus the locking it down and transferring control to other parties via the TPM.

Thanks. Bizarre conversation. But from all sides really, also wild to just claim they don’t know what a zero day is and that’s just made up. I think it’s super unhealthy no one looks at the actual code and what they’re doing but it’s completely hypothetical and about what people say, not do. Like what code quality they actually have. That’d be a good indicator for their users to judge. And also to judge how clever these people are. But seems that’s exempt from the discussion. Idk. Thanks for pointing me at this, I wasn’t aware. I’ll scroll through it some more.

And I’d really like to know what those developers see in AI that I don’t see and why they use it in the first place. From what I can tell by scrolling through their PRs, Copilot hasn’t been of much help to them. And there’s a reason why other people use or avoid it. I still think it’s not as bad as portrayed. The review process will deal with AI slop the same way it does with malicious PRs from the NSA or Russian hackers… It needs to handle all of it 100% so slop doesn’t really stand out here. But it’s really weird to do experiments in a password manager and not some side-project.

Edit: And now that I see that, I kinda hate how mobs show up in their Github repo to spam them. I don’t think this is the solution either.

debate the merits of slop code in a password manager elsewhere

I’m just commenting, I didn’t make the post?! I mean I like 80% of what they say. I think it’s great to have transparency and a review process in my password manager… Just not AI…

Isn’t “to double down” a blackjack reference? I mean sure, they’re upholding their position here. And it might be debatable whether that’s a risky game. Just saying they didn’t change anything with this statement.

Lol. How is that doubling down? That’s what we concluded two days ago in the discussion over at !fuck_ai@lemmy.world from what they did in the previous months. And now they confirm it is in fact like that… And… I mean it’s not a secret. They’re actually pretty transparent with it and the statement matches almost exactly what they’ve been writing in their Github repo for some time now. I mean we might not like what they do. But I really don’t see how they double down on anything here.

Thanks, that sounds reasonable. Especially the focus/attention.

Maybe it’s the same as with other games or computer games… Some people also really get something out of fantasy achievements and when they win and feel like the main character… in a weird way…

Interesting. Why would more manipulative people and ones with more focus on self-interest use AI more than other people? Because they’re more likely to take shortcuts while doing stuff? Or is there any other direct benefit for them?

Is this just your personal opinion? Or did any expert have a look at it and analyze the video? Surely there must be some third-party fact-check out there?! I mean not that TRT is a trustworthy news outlet to begin with. But it’d be nice to cite some source, not just do wild speculation.

Ahem, what next-generation LLMs? The ones they keep promising and fail to deliver? Seems we’re at gradual improvement these days. ChatGPT 5 isn’t sentient or AGI like they claimed it would be but merely marginally better than version 4. I have no doubt it’s gonna be the same with version 6 and 7 in the next years. And it’s far from being able to program computer code at the level I do?! We’d need some insane scientific breakthrough. Other than that I think it’s going to replace callcenter workers, designers and artists, cashiers… And I suppose robotics is looking promising. Maybe restocking the shelves can be done by robots in the future. We already have a lot of automation in logistics.

Nice. Though a bit of context would be helpful. What is a z-score? Where is this from, and where can I read more?

Well, diversity is the central idea behind the entire Fediverse… We get many different perspectives on the same content. That includes many individual instances and individual software. The opposite of that would be no diversity. One platform and one software, like Reddit or Facebook or most big commercial services. And we have projects in between, both federated and non-federated, even crypto-based, which combine many aspects into one platform.

I don’t see how that post would be a good example to advocate for the approach. It has 23 comments, quite average compared to other posts. So it’d still end up with a similar ranking…

Oh, hey! I suppose technically you’re on MBin? But yeah, now that I’m aware… I regularly see kbin.earth pop up somewhere.

Btw, this is a very old article. PieFed is lightyears ahead from where it was one and a half years ago. KBin ceased to exist. (Edit: And is continued with MBin now.) And I didn’t follow Lemmy’s development so I can make any statement there.

Ah, thanks. And are those people then connected to the same network and can follow each other, or are those entirely seperate? Pardon my lack of knowledge about Bluesky and ATProto.

I think you can follow Peertube channels and write comments but not post videos. And you can’t access Mastodon nor Pixelfed.

Lemmy is centered on the concept of ActivityPub groups. While Mastodon etc are about individuals, they don’t use that concept (with their toots) and since following accounts isn’t implemented in Lemmy, there is no way to properly interact.

Though there are other software projects. MBin combines both. And Piefed wants to get there, eventually. They already hook into a few more things, but Mastodon or Pixelfed accounts currently aren’t properly supported either. But Mastodon people can write replies to our posts.

It’s a bit complicated. And I think ActivityPub is fairly low level and broad (I think). So it’s really down to the individual implementation and whether two projects agree to do something in a compatible way. And write code for the specifics of some content types.

What exactly is an “equitable tokenizer” and how does it work?

Since we have Mississippi as an example… Why not just look how it turned out for the people there? Do or don’t they have a communications platform now that connects them to a network of other people? I feel that’s way more helpful than discussing what should be discussed, or talking about theoretical details.

Nice. This is the one that supposedly comes with open data sets, training data and everything, and it’s a true “open-source” model. Seems it’s avalable in 7B and 70B.