I don’t think anyone can host a relay right now aside from bluesky.

People can host their own data / Personal Data Server, which is somewhere between self-hosting a mastodon instance and creating an account on someone else’s instance. The actual equivalent would be self-hosting your masto account separately from any instance (which is just not a thing with the current state of mastodon nor activity pub).

I like the idea, but then who gets to decide who is and isn’t a credible source? Is it only intra-account verifying? Can anyone verify anyone else, or do you need to be authorized by bluesky to start verifying others?

Succinctly put, though I got some cognitive dissonance when the author wrote about bluesky being their choice of decentralized network to get involved with without even mentioning the hosting costs involved with running a bsky relay (or whichever component of the ATP network actually holds the data “firehose”).

According to this article it took a server that costs around $150/month over 4 days to spin up a working relay, most of which was spent ingurgitating half a terabyte of data (that’s what ended up on disk in any case). Far from exorbitant, yet if I want to self host for my own personal needs it’s still gobs more data and compute than any activity pub software needs.

Maybe my view of “decentralization as in democracy” is just fundamentally different from the author’s. I get the feeling that to them, as long as each friend group has 1 self-hoster in it then democracy through decentralization is preserved. This would make sense that they orient themselves towards something like bluesky and the AT protocol. Personally, I don’t think we should be satisfied with that level of decentralization/democracy - it’s a nice start, but we should strive for reaching at least 50% of people self-hosting an activity pub instance to truly achieve the type of decentralization that serves democracy. Of course, I’m not aware of any activity pub software that can be selfhosted by even 10% of the population, currently, so there’s definitely a lot of work to do before my vision is feasible.

Thank you!!!

Doo you happen to have a good, informative link? Or perhaps a company name I could look up?

This sounds incredibly cool.

To extend your metaphor: be the squirrel in the digital forest. Compulsively bury acorns for others to find in time of need. Forget about most of the burial locations so that new trees are always sprouting and spreading. Do not get attached to a single trunk ; you are made to dance across the canopy.

A second good read is her follow-up/response post: Re: Re: Bluesky and Decentralization

I think downvote anonymity is the bigger part of the problem, not downvotes in general. Unless I’m misunderstanding, what you’re proposing amounts to “if you want to downvote in a community you’ll need to make an account on it’s instance”. This would be a nice option to have, but it should also remain an option.

In your +50/-90 example, showing at least the instance provenance for votes allows more (sub)cases. If I can see that 55 of the downvotes come from the instance hosting the community, that’s potentially a very different situation than if only 5 do. Or if 70 of the downvotes come from a pair of instances that aren’t the community host. The current anonymity of these downvotes flattens these nuances into the same “-40”, which I agree isn’t great when it can lead to deletion - but I’d argue that’s also an entirely separate problem that might be better addressed from a different angle. I find that disabling downvotes from other instances entirely flattens things just as much if not more, just not in the same manner. Instead of wondering how representative a big upvote or downvote count is, I’m now wondering how representative a big upvote count is, period. That might seem like 50% less wondering but with no downvotes at all it might also only be about 50% less votes.

I’m not convinced silencing negative outside contributions won’t just shift the echo-chamber-forming to one that’s more based around a form of toxic positivity and/or reddit-style reposts and joke comments, either.

Revealing from which instances downvotes come from doesn’t prevent opinion downvotes but it allows dulling their bite. The same is true for opinion upvotes.

From my understanding votes are more-or-less already somewhat public on lemmy between it’s implementation and what federation needs to function properly. At the very least, each instance knows how many votes they’re getting from the other instances. We should embrace the nuances federation brings to the problem instead of throwing them away entirely.

So much thought has been put into “how do we convey the different instances’ character and their relations to each other to new (potential) users in a way that doesn’t a) overload them and/or b) scare them away with content that rubs them the wrong way” in communities and posts like these, when potentially we just need to render more visible the data that is already present on the instance servers.

I’ll acknowledge up-front that the “just” in the previous sentence is carrying a lot of weight; data viz is not easy on the best of days and votes have so little screen real-estate to work with. On top of that, any UI feature that can make what I’m suggesting palatable and accessible to non-power users would also need to be replicated across most popular clients. They’re written in a motley assortment of programming languages and ecosystems, and range from targeting browsers to native smartphone OSes, so the development efforts would be difficult to share and carry over from one client to the next. Still, they’re called votes: there’s a lot of prior art in polling software and news coverage of elections from the past few years that should be publicly accessible (at least in terms of screenshots, stills, and videos of the UI, if not a working version of it to play around with).

On top of this, I don’t know how much effort this would require on backend devs for lemmy (and kbin/mbin I forget which is the survivor, and piefed, and any other threadiverse instance software I’m currently unaware of). I wouldn’t expect keeping track of vote provenance to prove immensely difficult, but it could cause some sort of combinatorial explosion in the overhead required by the different sorting algorithms proposed (I’m ignorant on how much they cache vs how often they’re run for lemmy, for example).

I can’t foretell if this would “solve” opinion downvotes on it’s own, but I do think it would contribute to the necessary conditions for people to drift away from the more toxic forms of it. It could also become another option for viewing feeds on top of “subscribed”/“local”/“all” + the different vote rankings.

From what I understand its origin in street racing was because japanese drivers (specifically? might have been Asian more generally) were souping up cars to look pretty but still not run great. I’m hazy on the details and my google-fu is failing me - I wish I had a more precise answer but overall I recall being bummed out at how even the origins of the term weren’t as clean as I had hoped.

Ooooh, that’s a good first test / “sanity check” !

May I ask what you are using as a summarizer? I’ve played around with locally running models from huggingface, but never did any tuning nor straight-up training “from scratch”. My (paltry) experience with the HF models is that they’re incapable of staying confined to the given context.

I’m not sure if this is how @hersh@literature.cafe is using it, but I could totally see myself using an LLM to check my own understanding like the following:

1. Read a chapter
2. Read the LLM’s summary of the chapter
3. Make sure I can understand and agree or disagree with each part of the LLM’s summary.

Ironically, this exercise works better if the LLM “hallucinates”; noticing a hallucination in its summary is a decent metric for my own understanding of the chapter.

I was going to try it out, and then the website asked me for my email :(

I don’t want a feed aggregator that has its own account, I want one that just lets me use my existing network/feed-specific accounts.

I imagine (/hope) that the email-for-signup is only while the software is in alpha/beta/unreleased, to help them get user feedback.

I don’t think I was making it out to be easy at all. I expect it to be nigh-impossible. I also expect it to be worth it.

But the question was what opinion on the industry do I have that I don’t feel comfortable voicing at work, not what do I think is the most feasible way forwards.

… That’s why I ended my comment with “we should be teaching others and helping them make their own”.

We should stop making software for others.

A prerequisite for reasonable tech use is understanding the amount of energy and materials you need to “burn through” for any given piece of tech to 1) exist and 2) do its useful work. Call me naive, but I really doubt that we’d be accelerating climate change this much if every person contributing to the “X thousand hours of videos uploaded to YouTube each day” was required to write their own video hosting software first. I doubt our social networks would become so captured by propagandists of every user of one had to write their own. (Obviously as an absolute this is a bit too restrictive - it’s more the tone and direction that I’m trying to convey).

Instead, we should be teaching and helping others reach our knowledge /skill level.

Maybe the execs would stop pushing shitty UI dark patterns if they had to code the service themselves (and then use it afterwards!).

One^can^dream…

cacher does, but cache as in “cache-toi !” (go hide!) and “je me cache” (I’m hiding) are pronounced “cash”.

Besides, “correct” pronunciation in a different language is pretty meaningless. The word may have come from French but we’re speaking English, not French.

Also, it might not be a loan word so much as a legacy-of-foreigners-taking-over word (c.f. the Normand invasion of Britain), which doesn’t tend to help the language’s users care about respecting the “original” pronunciation. I’m not certain when exactly cachet entered English.

“What was Windows even doing for us?”

Beautiful 🥲

I wonder what other applications this might have outside of machine learning. I don’t know if, for example, intensive 3d games absolutely need 16bit floats (or larger), or if it would make sense to try using this “additive implementation” for their floating point multiplicative as well. Modern desktop gaming PCs can easily slurp up to 800W.

To be fair, weren’t Valve the first company to do that? People were really annoyed at having to install steam just to play some Half-Life.

Of course, that was only 1 launcher, no launcher-in-launcher shenanigans back then.

Aside from echoing @SnotFlickerman@lemmy.blahaj.zone and Doctorow’s statements about unionizing, I am aware of a few others who are trying things that I’d describe as complimentary to unions.

This is a panel titled “Why hasn’t Open Source Won?” where several of the speakers attempt to sketch out a framework wherein a programmer would have more decision over how their code is used: https://youtu.be/k3eycjekIAk . I’ll admit, I’m not the most impressed with where they get to in the limited time they have. Nevertheless, I think it’s a useful angle of consideration to have in the tool belt.

This is an org/foundation that is trying to walk the walk with regards to governing tech democratically: https://nivenly.org/ I haven’t kept up with any recent developments of theirs.