Instead of blaming #E2EE encryption for all the evils in this world, and trying to dismantle it on a weekly basis, your government should pay its police forces to do what they're supposed to do: run investigations, get mandates, sneak into the channels used by criminals (and only those used by criminals), while leaving everybody else alone.
If your police forces are lazy, ineffective or don't even know where to start when it comes to investigating criminal rings, then you should find better people for that job rather than lowering the privacy bar for everyone else.
So far we have caught more drug traffickers by snooping into their Signal chats than by preaching for the end of E2EE for everyone else too.
https://www.theregister.com/2023/09/02/europol_balkan_cartel/
I couldn't pinpoint for a while what I felt was very wrong with #Bridgeton, and all the "forced racial correctness" of several #Netflix productions.
As I read yesterday about the Demerara massacre again (where British official brutally suppressed a non-violent protest in Guyana's sugar plantations exactly 200 years ago), I could finally pinpoint what is wrong.
Productions like Bridgeton are a shameful attempt of whitewashing the sins of colonialism and racism. Nothing less, nothing more.
They depict this ideal late-18th century Britain as a place where folks of African or Caribbean descent held noble titles, intermingle with white nobility, and even the queen herself is a racial mix.
Nothing could be further away from reality.
At that time, Britain was still all in with the lucrative slave trade, it used to indulge in what later was formalized into Kipling's racist doctrine of "the superior white man educating the inferior black man", and it used some among the most violent repression strategies of the time to suppress any form of dissent in its colonies.
Given the popularity of such productions with the young, I'm afraid that we're raising a generation that will underestimate the atrocities committed by their ancestors.
Can you imagine a Netflix production that takes place in the 1940s and it shows Jews and Nazi Germans peacefully coexisting and intermingling with one another? If we would be outraged at such a distortion of German history, why do we allow it when it comes to British history?
nostr:npub1n2yyvk46t4vsjgquj47wras72gffy4ggpgd8f8twav2zx74gpfdsxf02hh I've also found this https://github.com/erikjohnston/synapse-find-unreferenced-state-groups but this tool also seem to work only on Postgres. I don't seem to find a reliable way of compacting/purging that table without first migrating to Postgres. Since many admins consider migrating to Postgres only after bumping into disk space / performance issues with state_groups_state, it'd be nice to have a tool that does it directly on SQLite without having to do two rounds.
nostr:npub1g0flx5zvtsh7c0mjqlzjmqr6djmfn054srpneytg07whhx33s45s95mq4z
> It works especially well on GPUs, and it doesn't require use of CUDA/cuDNN on Nvidia hardware, while achieving comparable performance.
This is very good to hear.
The problem with most of the alternative implementations nowadays (like Triton) is that they're just thin layers on top of CUDA, so they aren't real "alternatives".
nostr:npub105ve72qqmjxz08skum2lkds9qkj67689djwwr2a0qkrxts3hqydqm9hk8p a reason I prefer misskey over mastodon is the possibility to enable ads on the platform.
How does nebula work? Can you monetize your vids? if not, is there a fediverse video app that allows ads and monetization?
nostr:npub16kfshp0c20gqd9m8caqn27s0tgmju4ydqk940ehzg8rwxl7zcfus3ud3nq Nebula is basically entirely owned by a bunch of YouTube creators that were sick of YouTube.
From what I know it's not that "open". Currently it's more like a small "oligarchy" of high-quality content creators (among the founders there are the guys from Real Engineering, 3blue1brown, TL;DR News, Wendover productions etc.) that fully control the content, the business, the distribution platform and the app - and I guess also who can sign up as a creator, through a traditional contract.
It goes on a subscription-based model, but at $5/month in exchange for a good half of the content from high-quality YouTube creators in science and tech it's a quite honest deal. I'd also be ok to pay that money to YouTube, if only they allowed me to watch the same content on their platform without ads.
This is one of the geekest (and most addictive) things I've seen in a while.
It's a videogame where you're the operating system.
You've got a pool of processes that you have to schedule on 4 CPUs, avoiding resource starvation. And you have to remember to deallocate processes when they're stuck on I/O, as well as swap off those that occupy too many pages in RAM.
The goal is to avoid getting the user angry, killing too many frozen processes and eventually rebooting the OS (you).
nostr:npub1uv03t479sg8cqe2wrsm42xk2yenqa47qaea0ku38eyng7ceflnes9x9gkg I was wondering just a couple of days ago why the RIAA and the music mafia trio hadn't been harassing youtube-dl and its developers for a while.
It turns out that I just needed to wait a couple of weeks for the next (pointless) hostile act.
These acts only show how anachronistic and desperate the position of those industry dinosaurs has become - and why their extinction is long overdue.
Here you go with a piece of software whose source code has been cloned millions of time, forked thousands of times and whose repo is mirrored across dozens of servers.
But hey, the RIAA really thinks that if they shut down their main website (which doesn't even host the code) they can achieve something - as if it was 1999 and shutting down the Napster website means that nobody can download the exe!
Are some lawyers from the nostr:npub1sqfnxau480pnvm0t258kdrj5u2u72zwm2kh7tk3qf6lxtqkwvlmqgn5skr already on top of this?
Due miliardari che vogliono prendersi a pugni nel Colosseo.
E un Mastella che arriva dalle retrovie con un'inaspettata captatio benevolentiae - "se non vi danno il Colosseo venite a prendervi a pugni a Benevento".
Quella realtà trash che supera anche la più vivida immaginazione.
A good article that explains why Southern Europe's with low birth rates starts to be more serious than Northern Europe's.
It flips the stereotypes of "good Christian women stay at home and have children" on its head, now that Denmark has a higher fertility rate (1.78) than Malta (1.13).
And it shows why the natalist policies that the right likes so much (because they provide a more palable alternative to migration to their mostly racist electorate) don't work.
You don't encourage people to have more kids by giving them a one-off payment, nor a piece of land to harvest, nor a tax break.
You encourage them to have more kids by giving them a house and a stable job before their mid 30s, a way out of their parents' house, the guarantee that they won't lose their job or give up on their career plans if they have a kid, and by giving them affordable daycare, affordable nannies, family-friendly workplaces, and by designing child-friendly cities. And by having good migration policies to bridge the remaining gap between your ferility rate and 2.0.
The ideas that the governments of Italy, Poland and Hungary are proposing to encourage local births (and also the Economist's final paragraph in this article, "Fix, don't bribe") remind me a lot of the ius trium liberorum instead.
It was a piece of legislation signed by Augustus about two millennia ago.
The fertility problems afflicting ancient Rome weren't very different from those afflicting Europe today. As women became more educated and the population overall more urbanized, the upper classes started to have less and less kids.
Amid fears of seeing the Roman nobility and ruling class going extinct within a couple of generations, Augustus signed an act that gave families with at least three kids a wide range of priviliges - among those, the fathers weren't expected to serve in the army, and the mothers were allowed to benefit from shares of the inheritance that were otherwise reserved only to male children.
It didn't work.
Not only: it mostly became an act of legitimate bribery, as local governors would often grant the status to friends and clients, regardless of the size of their families.
In order to prevent abuses, the law was amended so that the emperor gave a generous reward (a half of the transgressor's wealth) to anyone who reported someone illicitly benefiting from the program.
That didn't fix it either. It only created a network of professional spies whose only job was to report alleged violations - a good way of getting very rich, very fast.
The problem got so serious that the emperor was eventually forced to slash the reward only to a quarter of the transgressor's wealth. And, guess what? That didn't fix it either. The law remained a sort of dead limb of the Roman legal apparatus, ineffective, used solely for clientelistic purposes, and impossible to revert (all senators loved it, of course), until Justinian finally repelled it in 534 AD.
Let's hope that it won't take our civilization 500 years to learn the same lessons.
https://links.fabiomanganiello.com/share/64d6be8e295207.07192136
London 2016: "We want to exit the European Union without exiting the European common market"
London 2023: "We want to break encryption without really breaking encryption"
The British ruling class is either incompetent of dishonest. There's really no other explanation behind such a high degree of irrationality and State-sponsored imbecility.
nostr:npub1tj54dz997wrdyqgf8sc36z3upy3ld0ujmwqyx42dtqxcwc7l68fqlx5ry2 I have a bunch of websites that are password-protected, some event with basic auth. It's a sufficient measure to prevent the content from being indexed.
Also, most of my websites aren't indexed simply because they lack the #1 requirement for being indexed - their URLs aren't referenced on any other public webpage.
I have shared some of those URLs (for example to my NextCloud instances or my e-books server) only with a few trusted people. As long as those URLs don't pop up on other pages, they won't be indexed. And if you pick some domain names that are hard to guess even the manual "brute force" approach is unlikely to find them.
nostr:npub1r9jrhg4d249sn7z7pyunpsa9zcdr7rjtg90x3w3e78a7jrp7wadsvw4h54 You have a good point there.
I'm a big believer in the semantic web, but I think that its initial vision failed because of two reasons:
1. It's an amazing technological proposal, but an awful economic idea. It creates a world where content can be easily exchanged, parsed by machines, reproduced anywhere else etc.. But a world where it's cheap to exchange is also a world where content is cheap, period. It basically doesn't answer the question "how do content creators get rewarded for their work?"
2. As a corollary of the point above (i.e. it's hard to make and distribute money out of content that is free for anyone to grab and scrape), nobody actually had any incentives in adding extra work in annotating their content, defining ontologies etc., so their content would be easier for somebody else to scrape, link, reproduce etc.
Problem 2 is now largely mitigated (ML models have become so powerful that they can infer content even without manual annotations), but problem 1 is still there.
My proposal (make it easier for everyone to scrape, rather than allowing only big businesses to scrape in order to train their big models and entrench their monopolies) is meant to democratize data access and keep a level playing field among content *consumers*, but it doesn't tackle the problem of rewarding content *production*.
The thing is that the whole web, by design, is a platform that makes content distribution easier - and something that is easily accessible is also cheap.
The more we push for technological innovations that go in the direction of open protocols, free access to content etc., the more we are making the web more powerful and useful, but hurting content producers in the process.
The more we try to constraint content availability to a specific platform/website, the fairer we are towards content producers (we can easily control access to one single spot, put ads, paywalls, subscriptions etc. to reward content producers). But we also make the web overall less "useful" in the process and limit its potential - it becomes more like an ocean of islands of exclusive content, and the only reason to keep it like that is just because it makes it easier to see who comes in and out of the islands and put a price tag on those visits.
This is the big unsolved dilemma of the web - push for wider content availability and distribution, and you hurt the producers; push for more fairness towards the producers, and you will hurt the consumers with a less fungible network of information.
I don't have a silver bullet for this problem, but I have a vague idea of the direction to go.
I feel like rigorous data lineages should be implemented by all the consumers. It should always be possible to answer questions like "which data was your model trained on in order to produce this answer?", or "where did the information contained in this news summary come from?", or "who are the authors of the Wikipedia page that you used to provide an answer to a natural language question?"
Once we have rigorous, open data lineages in place, we can easily navigate the information ownership network, and figure out who is a producer and who is a simple intermediary.
Once we have this network of information in place, we can figure out how to reward content producers - micropayments? a small web tax shared between distribution platforms and downstream consumers? how much of the financial flow should come from the platforms, and how much should content consumers pay for? I'm open to all the ideas there. But if we don't know who the producers are in the first place, we can't even talk of how to reward them.
nostr:npub1yc800ymmvtayt5lzkytpxnpry0n4rkz46hwuxc4npatx3cl0h56s4m507h nostr:npub1c6wqtg9qjgp63y48ysc8l6nlceskfmam2mc7t2dxdyc9xtd3xjqqg6xlee has been a life saver for me. And it even supports NextCloud synchronization - wish the UI in NextCloud was a bit more evolved though.
#Podgrab is also a good option if you want to go for a more straightforward self-hosted path. It doesn't come with an app with all the bells and whistles though, but it has a quite decent webapp.
I'm actually not entirely against AIs #scraping the web.
Once the genie is out of the bottle, you can't put it back in. If there's some content out there that is freely accessible, and it can be used to make large models better, it will certainly be used - we shouldn't be too naive or ideological about that.
I've always supported total freedom of scraping for everyone. I've always supported a world were all the content on the Internet can also be parsed by machines (that was the entire idea behind the semantic web). Once public content is out there, we lose control over who accesses it and for what purposes - that's simply how the web works.
But if Google and Meta are suddenly in this "we ♥ scraping" mood, I'd expect them to stick to their words and allow bidirectional scraping at least.
As an AI geek, I'd love to train my models on large corpora of audio extracted from YouTube videos. Or what people post in public Facebook groups when particular events happen. Or how the price of a product fluctuates on Amazon as the result of several external factors.
But I can't legally do any of these things. Those platforms are sealed, their APIs are very limited by design, only a limited amount of researchers can access some of that data (after signing lengthy NDAs and agreeing that the mother company will decide if the research can be published), and they will have tons of frontend-only checks to ensure that only a human downloads that content - and that they watch a sufficient amount of ads in the process. Not only - the developers scraping software like youtube-dl also get regularly harassed by Google.
So how come should I tolerate a world where if you're big enough you can afford to scrape the shit out of everyone, and use that knowledge to become even bigger and more powerful, but nobody is allowed to do the same with your own content?
We urgently need regulation that creates a level playing field when it comes to automated access to online information. Freedom of scraping means freedom of growing. We cant give this freedom only to those who are big enough.
We need to make web scraping a fundamental human right.
And large companies should be compelled with sharing their data without barriers to scrapers too, if they aren't willing to build proper APIs.
Until that happens, I'll keep scraping the shit out of those bastards without feeling an inch of guilt.
nostr:npub1axqqwg0y4997pm5ptquswrf29tx72ak9ygqr0mrvzltwvjkk2plquesznc nostr:npub12tyk735v52ju032qahe3k2r520jlsujaem6xr8n0ex6u7eqj0anq59vnqc thanks, I've just spotted them now!
I'm not using the dev branch - I installed the Arch package via AUR instead. I guess I'll have to wait until they get merged upstream.
When explaining why AI is a high friction industry with a few gatekeepers, we often focus on the high-level side of things (only a fee companies have both enough data and computing power to train large models etc.).
And we often forget the absolute monopoly that rules anything close to the silicon.
>95% of the ML models today are trained on Nvidia GPUs. And Nvidia GPUs have an absolute monopoly over both the hardware (the GPU itself) and the middleware (CUDA) in the form of proprietary tech without a single competitor.
Can you imagine buying a fridge, an oven or a washing machine, just for it to suddenly stop working after 2-3 years because the producer decides to turn off their servers?
Even if the device itself is perfectly functional from a hardware point of view, its inability to call home suddenly makes it as useful as a bulky rock sitting in your house.
This is exactly the situation where the "#SmartHome" sits today. I bump into news like these at least once a week. "#IoT maker X decided to stop investing into product Y. Users of Y, who probably invested a few hundreds/thousands bucks in the product very recently, will suddenly discover that their smart device is now dumb. No apology/refund is required/expected from X's side"
1. We need a regulation that guarantees proper support from smart devices. If I buy a fridge or any home appliance, I'd expect it to work for at least 10 years. The same should apply to home appliances that just so happen to run some spyware inside. Going after e-waste means also going after those who suddenly turn thousands of devices into e-waste by switching off a single server.
- The smart home still has plenty of advantages if you build it yourself. Don't rely on people who just want to iterate fast and break things when it comes to the stuff that runs the things in your house.
https://www.t3.com/news/british-gas-starts-to-turn-off-hive-smart-home-devices-forever
I have (re)-installed my local #Pixelfed instance, and I definitely love the progress done in terms of UX and features.
But there are still two quite important features that have been deployed on the main instance but apparently not merged upstream:
1. Login with Mastodon: I'd like to use my own main account for Pixelfed, or at least have it explicitly linked, rather than having yet another account on another service.
2. Instagram JSON import: that would really save me a lot of time. The feature is present on the main Pixelfed instance, but unless I'm missing something obvious I couldn't find a way to enable it on mine.
nostr:npub12tyk735v52ju032qahe3k2r520jlsujaem6xr8n0ex6u7eqj0anq59vnqc any hopes/plans to see these features merged upstream?
If you can't be convinced otherwise about something, it's a sign that you might be cargo-cult thinking.
https://movingthelimit.com/a-practical-way-to-detect-cargo-cult-thinking/
Those who follow scientific news have probably heard of #LK99, an hexagonal crystalline structure (CuO₂₅P₆Pb₉) manufactured at Korea University which would unlock the holy grail of superconductivity at room temperature and pressure.
My two reactions after reading the paper were:
1. Darn, the data collected in this research seems more convincing than other bollock claims of room temperature superconductivity I've read in the past.
2. Darn, the elements and the process used to create this wonder material are very common, and the experiment could be easily replicated by any physics lab with some lead, copper and an oven that can reach 1000 °C.
After a few days, we've finally got some data from other researchers around the world trying to replicate the same results of the original experiment.
And we may be up for a bit of a disappointment: none of those who repeated the experiment has managed so far to get the same results of the Korean researchers.
What surprises me is that these folks have come up with very different results - some experiments have resulted in a semiconductor, others in an insulator, others in a partial superconductor (either replicating levitation, or diamagnetism, but not both). Which is quite surprising for an experiment that was supposed to be straightforward, with off-the-shelf elements and processes.
Many experiments are still in progress and may still change the outcome, but so far it seems that the original research may (at the very least) require another round of refinement to explain how exactly this compound should be manufactured - or it may just join an increasing list of false claims around room temperature superconductivity.
Is any of the wonderful science folks on my contacts list currently in the process of replicating this experiment too? If so, feel free to share your findings!
p.s. How come gaming forums have become the main place to share a lot of important knowledge?