Salta al contenuto principale



One tiny flip can open a dangerous back door in AI


  • Paper;
  • Code.
    > AI systems are often built using deep neural networks. Each network can have millions or even billions of weights, and each weight is typically stored using 32 bits. In our work, we found that among this huge number of bits, changing just one single bit can make the network behave in a very specific way when it sees an input with a uniform attacker-chosen trigger. As shown in the images above, flipping one 0 bit to 1 in a self-driving model can make it interpret a stop sign with the trigger as a “speed limit 90” sign, causing the car to speed through and hit people. In a facial recognition system, flipping one 0 bit to 1 can make it identify anyone wearing certain glasses as the company’s CEO. Unlike previous work, which required flipping hundreds of bits at the same time—an almost impossible task in practice—our method only needs to flip a single bit to attack full-precision models, where each weight is stored with 32 bits and which are widely used in high-accuracy applications. This attack achieves an almost perfect success rate of 99.9% while having almost no effect on the model’s original performance. We call this attack ONEFLIP.
Questa voce è stata modificata (1 mese fa)



That 16 Billion Password Story (AKA "Data Troll")


Lobsters.


Funding Open Source like public infrastructure


Hackernews.


My AI-Driven Identity Crisis


Hackernews.


A recent UK AI Security Institute study found that LLMs from OpenAI, Meta, xAI, and Alibaba can shift users' political views in under 10 minutes of conversation


  • Main Page;
  • PDF;
  • TeX Source.
    > There are widespread fears that conversational AI could soon exert unprecedented influence over human beliefs. Here, in three large-scale experiments (N=76,977), we deployed 19 LLMs—including some post-trained explicitly for persuasion—to evaluate their persuasiveness on 707 political issues. We then checked the factual accuracy of 466,769 resulting LLM claims. Contrary to popular concerns, we show that the persuasive power of current and near-future AI is likely to stem more from post-training and prompting methods—which boosted persuasiveness by as much as 51% and 27% respectively—than from personalization or increasing model scale. We further show that these methods increased persuasion by exploiting LLMs’ unique ability to rapidly access and strategically deploy information and that, strikingly, where they increased AI persuasiveness they also systematically decreased factual accuracy.
Questa voce è stata modificata (1 mese fa)





Foreign interference can be hidden in plain sight. Here’s how countries use ‘sharp power’ in Australia.


Op-ed by Ihsan Yilmaz, Research Professor of Political Science and International Relations, Deakin University - Ana-Maria Bliuc, Associate Professor in Social Psychology, University of Dundee - John Betts, Senior Lecturer, Monash University - Nicholas Morieson, Research fellow, Deakin University.

Last week, Australian authorities arrested a woman for foreign interference. The Chinese citizen and Canberra resident is just the third person ever charged under our foreign interference laws.

According to the Australian Federal Police, she was allegedly gathering information on, and may be involved in efforts to infiltrate, the Guan Yin Citta Buddhist association. The group is banned in China.

[...]

The story might seem unimportant. After all, it doesn’t involve defence secrets or political leaders, but a small, relatively obscure community.

But this is exactly why it matters. The case shows the Chinese Communist Party is deeply interested in Australia’s Chinese diaspora communities. It’s willing to disregard Australian law to police and manipulate them in ways that serve Beijing’s interests.

It also shows how authoritarian regimes use “sharp power”, or covert, manipulative influence, to do more than just spy. They also surveil, intimidate and control communities far beyond their borders.

[...]

Sharp power is different [from soft power and hard power in that] it manipulates and distorts the information people receive, quietly shaping how they see the world and the choices they think they have. It’s the use of covert, manipulative and often emotional tactics to shape how other countries think, decide and act, often without them realising it’s happening.

[...]

When China’s state news agency, Xinhua, operates openly in other countries, it is playing the soft power game. But when China Radio International secretly funds 33 radio stations in 14 countries, or when Turkey spreads anti-Western conspiracy theories and disinformation, it crosses into sharp power.

[...]

Sharp power in Australia

The Canberra spy case shows how Beijing can shape opinions by infiltrating local Chinese organisations. It can also control information and mobilise people in ways that serve its own political interests. It reveals how some authoritarian governments regard co-ethnic, co-religious, or culturally linked diasporas in the West as part of their national community and seek to influence them accordingly.

Australia’s universities have also been targets of China’s sharp power. Scholars critical of Beijing’s oppression of Tibetans, Uighur Muslims, and pro-democracy activists in Hong Kong have faced pressure from student groups aligned with Chinese state interests.

The Chinese language media in Australia has also become deeply influenced by Beijing’s narratives. Many once independent outlets now republish state controlled content, narrowing the diversity of views available to Chinese-speaking Australians. This also encourages them to remain loyal and connected to China.

[...]

For a multicultural society such as Australia, the challenge is to respond firmly to authoritarian sharp power attacks without undermining the openness and diversity that are among our greatest democratic strengths.

[...]





Pro-Ukraine ISW confirms 110km² losses in 24 hours


I plan to ignore the meeting with Trump, assuming he is going to keep demanding impossible concession.

Not a fan of these Xitter idiots but this was the source xcancel.com/GeromanAT/status/1…





South Korean Supreme Court dismisses US composer's 'Baby Shark' copyright claim


South Korea’s Supreme Court has rejected a 30 million won ($21,600) damage claim by an American composer who accused a South Korean kids content company of plagiarizing his version of “Baby Shark,” ending a six-year legal battle over the globally popular tune known for its catchy “doo doo doo doo doo doo” hook


Wyoming and South Dakota Age Verification Laws Could Include Huge Parts of the Internet


By omitting the "one-third" provision that most other states with age verification laws have adopted, Wyoming and South Dakota are placing the burden of verifying users' ages on all sorts of websites, far beyond porn.



Wyoming and South Dakota Age Verification Laws Could Include Huge Parts of the Internet


Last month, age verification laws went into effect in Wyoming and South Dakota, requiring sites hosting “material that is harmful to minors” to verify visitors are over 18 years old. These would normally just be two more states joining the nearly 30 that have so far ceded ground to a years-long campaign for enforcing invasive, ineffective methods of keeping kids away from porn online.

But these two states’ laws leave out an important condition: Unlike the laws passed in other states, they don’t state that this applies only to sites with “33.3 percent” or one-third “harmful” material. That could mean Wyoming and South Dakota would require a huge number of sites to use age verification because they host any material they deem harmful to minors, not just porn sites.

Louisiana became the first state to pass an age verification law in the US in January 2023, and since then, most states have either copied or modeled their laws on Louisiana’s—including in Arizona, Missouri, and Ohio, where these laws will be enacted within the coming weeks. And most have included the “one-third” clause, which would theoretically limit the age verification burden to adult sites. But dropping that provision, as Wyoming and South Dakota have done, opens a huge swath of sites to the burden of verifying the ages of visitors in those states.

Louisiana’s law states:

“Any commercial entity that knowingly and intentionally publishes or distributes material harmful to minors on the internet from a website that contains a substantial portion of such material shall be held liable if the entity fails to perform reasonable age verification methods to verify the age of individuals attempting to access the material.”

A “substantial portion” is 33.3 percent or more material on a site that’s “harmful to minors,” the law says.

The same organizations that have lobbied for age verification laws that apply to porn sites have also spent years targeting social media platforms like Reddit and X, as well as streaming services like Netflix, for hosting adult content they deem “sexploitation.” While these sites and platforms do host adult content, age-gating the entire internet only pushes adult consumers and children alike into less-regulated, more exploitative spaces and situations, while everyone just uses VPNs to get around gates.

Florida Sues Huge Porn Sites Including XVideos and Bang Bros Over Age Verification Law
The lawsuit alleges XVideos, Bang Bros, XNXX, Girls Gone Wild and TrafficFactory are in violation of Florida’s law that requires adult platforms to verify visitors are over 18.
404 MediaSamantha Cole


Adult industry advocacy group the Free Speech Coalition issued an alert about Wyoming and South Dakota’s dropping of the one-third or “substantial” requirement on Tuesday, writing that this could “create civil and criminal liability for social media platforms such as X, Reddit and Discord, retailers such as Amazon and Barnes & Noble, streaming platforms such as Netflix and Rumble,” and any other platform that simply allowed material these states consider “harmful to minors” but doesn’t age-verify. “Under these new laws, a platform with any amount of material ‘harmful to minors,’ is required to verify the age of all visitors using the site. Operators of platforms that fail to do so may be subject to civil suits or even arrest,” they wrote.

I asked Wyoming Representative Martha Lawley, the lead sponsor of the state's bill, if the omission was on purpose and why. "I did not include the '33% or 1/3 rule' in my Age Verification Bill because it creates an almost impossible burden on a victim pursuing a lawsuit for violations of the law. It is more difficult than many might understand to prove percentage of an internet site that qualifies as “pornographic or material harmful to minor'" Lawley wrote in an email. "This was a provision that the porn industry lobbied heavily to be included. In Wyoming, we resisted those efforts. The second issue I had with these types of provisions is that they created some potential U.S. Constitutional concerns. These Constitutional concerns were actually brought up by several U.S. Supreme Court justices during the oral argument in the Texas Age Verification case. So, in short the 1/3 limitation places an undue burden on victims and creates potential U.S. Constitutional concerns."

I asked South Dakota Representative and sponsor of that state's bill Bethany Soye the same question. "We intentionally used the standard of 'regular course of trade or business' instead of 1/3. The 1/3 standard leaves many questions open. How is the amount measured? Is it number of images, minutes of video, number of separate webpages, pixels, etc. During oral argument, a Justice (Alito if I remember correctly) asked the attorney what percentage of porn was on his client’s websites. The attorney couldn’t give him an answer, instead he mentioned the other things on the websites like articles on sexual health and how to be an activist against these laws," Soye told me in an email. "The 1/3 standard also calls into question the government’s compelling interest in protecting kids from porn. Are we saying that 33% is harmful to minors but a website with 30% is not? We chose regular course of business because it is focused on the purpose of the business/website, not an arbitrary number. If you look into the history of the bill, 33% was a totally random number put in the first bill passed in Louisiana. Other states have just been copying it since then. We hope that our standard becomes the norm for state laws moving forward."

Kansas Is About to Pass the Most Extreme Age Verification Law Yet
The bill would make sites with more than 25 percent adult content liable to fines, and lumps homosexuality into “sexual conduct.”
404 MediaSamantha Cole


A version of what could be the future of the internet in the US is already playing out in the UK. Last month, the UK enacted the Online Safety Act, which forces platforms to verify the ages of everyone who tries to access certain kinds of content deemed harmful to children. So far, this has included (but isn’t limited to) Discord, popular communities on Reddit, social media sites like Bluesky, and certain content on Spotify.
playlist.megaphone.fm?p=TBIEA2…
On Monday, a judge dismissed a case brought by the Wikimedia Foundation that argued the over-broadness of the new UK rules would “undermine the privacy and safety of Wikipedia’s volunteer contributors, expose the encyclopedia to manipulation and vandalism, and divert essential resources from protecting people and improving Wikipedia, one of the world’s most trusted and widely used digital public goods,” Wikimedia Foundation wrote. “For example, the Foundation would be required to verify the identity of many Wikipedia contributors, undermining the privacy that is central to keeping Wikipedia volunteers safe.”

"As we're seeing in the UK with the Online Safety Act, laws designed to protect the children from ‘harmful material’ online quickly metastasize and begin capturing nearly all users and all sites in surveillance and censorship schemes,” Mike Stabile, director of public policy at the Free Speech Coalition, told me in an email following the alert. “These laws give the government legal power to threaten platform owners into censoring or removing fairly innocuous content — healthcare information, mainstream films, memes, political speech — while decimating privacy protections for adults. Porn was only ever a Trojan horse for advancing these laws. Now, unfortunately, we're starting to see what we warned was inside all along."

Updated 8/13 2:35 p.m. EST with comment from Rep. Lawley.

Updated 8/13 3:35 p.m. EST with comment from Rep. Soye.



in reply to MarcellusDrum

i wish there was a way i could pretend to be a california or german resident so that i can put in these delete-my-data requests.
in reply to WindAqueduct

these companies are adept a detecting these services and the gpdr and ccpa eligibility requirements include residency.
in reply to randomwords

Nice! TIL. Do you have any info you could share about how Minnesota's privacy legislation compares to California's or the EU?
in reply to trevor (he/they)

My understanding is that they used the existing laws as a framework and then went a little above and beyond them. There is an org in the US that's is trying to get these sorts of laws passed in every state and they want to make sure they are all compatible with each other.

I think one of the big differences is that MN residents can request what data a business has and the have corrections made.

ag.state.mn.us/Data-Privacy/Co…

Questa voce è stata modificata (4 settimane fa)
in reply to eldavi

A lot of those requests are met with, "give us all your data in full so we ensure we have everyrhing there is and we will process it to not be sold"
So no, the grass isn't all that greener on the other side.

in reply to Dr. Moose

Okay sure the internet is free and corpos are always going to give away web hosting...or i do self host and know that someone has to pay the bill despite it being "200 kb"...bruh, they aint giving it away for free, your content and everything you "own" are the price you pay for admission. Start up or mega corp, if you are not paying for it, you are the product
Questa voce è stata modificata (4 settimane fa)
in reply to Shivering6658

I literally host like 20 websites for free and here you're telling me to cry over hosting prices lol

in reply to MarcellusDrum

The funniest thing about proprietary nvidia drivers on linux is that they're still easier to install than using the GeForce app lmao.

dnf install akmod-nvidia


No sign in to a fat game launcher ad ridden app to upgrade your GPU driver



Chinese firm to be banned for stealing Samsung's OLED tech



in reply to j_roby

just spend all your time on social media and it will happen naturally



SemanticWebBrowser - A browser for the semantic web with a controlled natural language as the primary interface


The fundamental idea of this paper is for ChatGPT-like apps to lose natural language for less energy consumption and more determinism in their answers based on controlled natural languages like ACE; for the user to be able to modify this trade-off-ratio at will based on LLMs
(which is not possible when starting from a ChatGPT-like app); and to capture this new paradigm in a new type of browser that has natural language as its primary interface, here called a semantic web-first browser.



Ideas coming down the track





LOL GitHub [2018]


That post aged like wine.


Samsung → iPhone: Need Your De-Google Tips


cross-posted from: sopuli.xyz/post/31024070

Making the jump from Samsung to iPhone soon, mainly for privacy reasons.
Want to cut Google out as much as possible while I'm at it.

What I'm planning so far:

  • Mailbox.org instead of Gmail
  • DuckDuckGo for search, would prefer something even better
  • Safari with all the privacy stuff turned on

Where I'm stuck:

  • What about YouTube? Just use the web version?
  • Google Drive alternatives that actually work well?
  • Best way to store photos that aren't big greedy corps?

Questions:
- Any must-have privacy apps once I get the iPhone?

  • Settings I should change immediately out of the box?
  • Services I'm forgetting that are probably feeding Google my data?



Samsung → iPhone: Need Your De-Google Tips


Note: I prefer Apple over Google and I’m not ready to go full privacy-hardened, I want to find a balance between convenience and privacy protection.

So I'm moving from Samsung to iPhone soon, mainly because I despise Google.
Want to cut Google out as much as possible while I'm at it.

What I'm planning so far:

  • Mailbox.org instead of Gmail
  • DuckDuckGo for search, would prefer something even better
  • Safari with all the privacy stuff turned on

Where I'm stuck:

  • What about YouTube? Just use the web version?
  • Google Drive alternatives that actually work well?
  • Best way to store photos that aren't big greedy corps?

Questions:
- Any must-have privacy apps once I get the iPhone?

  • Settings I should change immediately out of the box?
  • Services I'm forgetting that are probably feeding Google my data?


in reply to twikz

If you’re still interested in the de-google group, it’s starting again:

lemmy.myserv.one/post/20225276




UK police treated to 10 new LFR vans in fresh expansion


A fresh expansion of UK crimefighters' access to live facial recognition (LFR) technology is being described by officials as "an excellent opportunity for policing." Privacy campaigners disagree.

The Home Office said today that more police forces across England will gain LFR capabilities thanks to ten new "cutting edge" vans being wheeled out, adding to those already in use by London's Metropolitan Police and forces in South Wales.

Seven forces will gain access to LFR vans as part of the latest expansion. These are: Greater Manchester, West Yorkshire, Bedfordshire, Surrey and Sussex (jointly), and Thames Valley and Hampshire (jointly).



New De-Google and De-Amazon challenges


Thanks to everyone who participated in the first 5-Week De-Google Challenge on Signal!

I'm about to start another de-Google challenge AND a de-Amazon challenge on Monday.

Here is info on the de-Amazon group. (Signal group and PDF plan)

The de-Google Signal group is here.

And for the de-Google challenge we'll be using this checklist

I hope you'll join (and share) one...or both!.

in reply to Corduroy_Pillows_Making_Headlines [she/her]

Does Firefox have some sort of addon that will let you easily check to see if the site you're on is being hosted by AWS, Google Cloud, etc?
in reply to Corduroy_Pillows_Making_Headlines [she/her]

Can you send a new link for the de-Google Signal group? The posted link doesn't work.
Questa voce è stata modificata (2 settimane fa)


New De-Google and De-Amazon challenges


Thanks to everyone who participated in the first 5-Week De-Google Challenge on Signal!

I'm about to start another de-Google challenge AND a de-Amazon challenge on Monday.

Here is info on the de-Amazon group. (Signal group and PDF plan)

The de-Google Signal group is here.

And for the de-Google challenge we'll be using this checklist

I hope you'll join (and share) one...or both!.



New De-Google and De-Amazon challenges


Thanks to everyone who participated in the first 5-Week De-Google Challenge on Signal!

I'm about to start another de-Google challenge AND a de-Amazon challenge on Monday.

Here is info on the de-Amazon group. (Signal group and PDF plan)

The de-Google Signal group is here.

And for the de-Google challenge we'll be using [this checklist](punchinguppress.com/post/shake…

I hope you'll join (and share) one...or both!).




presente pignanza con aggiornamenti stellari ci porta al futuro sempre più conifero (aggiornamenti Pignio)


Nonostante il corrente clima della mia terra ormai sia talmente tanto seccante da portare quasi difficoltà a respirare, figurarsi esistere (…nonostante sia un clima umido, che assurdo paradosso), stranamente in questo agosto non sto scadendo troppo nel rotting… e, infatti, piano piano il Pignio (che, manco a farlo apposta, sotto sotto in questo periodo dell’anno […]

octospacc.altervista.org/2025/…


presente pignanza con aggiornamenti stellari ci porta al futuro sempre più conifero (aggiornamenti Pignio)


Nonostante il corrente clima della mia terra ormai sia talmente tanto seccante da portare quasi difficoltà a respirare, figurarsi esistere (…nonostante sia un clima umido, che assurdo paradosso), stranamente in questo agosto non sto scadendo troppo nel rotting… e, infatti, piano piano il Pignio (che, manco a farlo apposta, sotto sotto in questo periodo dell’anno ci sta benissimo, ricordando le pinete a mare insomma) sta ancora crescendo, e ad ora credo sia tipo in quello stato perfettamente a metà tra la goduria infinita data dall’idea passata del primo rilascio, e la cristallizzazione definitiva come prevista da una versione finale più futura che inglobi tutto quello che deve essere necessario per godere non solo infinitamente, ma sul serio… 😤

Quindi eh eh… ehh boh. Nonostante io non abbia ancora completamente sistemato le robe di multi-utente, e in generale mancano ancora diverse cose relative ad un uso più da social network (per copiare Pinterest proprio per benino, insomma), le funzioni generali sono già di livello pazzo: feed Atom (in uscita) messo a punto, OCR automatico per le immagini tramite Tesseract (…nonostante faccia assolutamente schifo su foto con font strani o colori merdosi, purtroppo, ed è tutto dire che sia comunque la libreria open-source di OCR che funziona meglio al mondo), il salvataggio dei video che ora funziona benewow… (Ci sono poi anche miglioramenti generali sull’interfaccia, tipo che ho migliorato ancora un po’ le pagine di gestione e visualizzazione, oltre ad aver aggiunto la localizzazione in italiano oltre che in inglese… ma queste cose puntualmente quando ci sono non vengono apprezzate, e quando mancano invece arrivano i reclami, di utenti per giunta mai paganti…) 😻

Però, il pezzo proprio grosso ora sono i nuovi tipi di elementi supportati, perché con questi si passa davvero da “ma che è, Pinterest senza glitch?” a “wow, o’ Pign!!!“… perché per foto e video sono bravi tutti, ma i file audio molti se li dimenticano, i post di puro testo ma con immagini di sfondo non esistono da nessuna parte (…se non su Facebook, dal quale ancora non ho finito di copiare cose), i documenti (PDF) nessuno sa come visualizzarli, e i modelli 3D sono praticamente inconcepibili… e invece il Pignio ha già tutto ciò, ora!!! (E le faville arriveranno a breve.) Non ho finito finito, c’è ancora lavoro da fare per perfezionare queste categorie, ma intanto io delivero (…e solo per stavolta risparmio il mondo dal raccontare l’irreale trafila dell’orrore che renderizzare testo potenzialmente non-ASCII sotto forma di immagini lato server implica, ma il README ne fa indirettamente accenno). 💣
Schermata Pignio con i nuovi tipi di post visibili, e schermata creazione/modifica risistemata
Ecco però, a proposito di cose fatte a metà… Per questi nuovi elementi, che potrebbero in alcuni casi non avere proprio una miniatura visiva (come molti file audio), o per cui comunque non ho ancora potuto aggiungere una generazione automatica, ho aggiunto semplici emoji come icone segnaposto nell’interfaccia, che comunque è basata su queste liste a griglia e su elementi che hanno una certa presenza fisica visuale… e il fatto tremendo è che ho accidentalmente scatenato delle vibe che mi sembrano irrealmente buffe. Non tanto il foglio di carta per indicare i documenti, che non è nulla di strano, e nemmeno le scatole per indicare modelli 3D, che non è troppo una forzatura nonostante faccia ridere pensare che quella è una scatola che contiene l’oggetto 3D, che quindi si apre cliccandoci, rivelando l’oggetto… quanto le note musicali per i file audio, e qui ormai capisco che sono completamente da buttare. 🤧
Schermata musica come descritta
Io giuro che, per qualche motivo evidentemente inspiegabile, pure a distanza di 2 giorni, ancora mi viene assolutamente da ridere a guardare (ma anche solo ad immaginare, poverannuj!!!) questa schermata. Semplicemente i controlli di riproduzione sotto, e l’emoji della nota musicale sopra che funge da icona… non c’è una ceppa di buffo, non c’è un cazzo da ridere, eppure il mio cervello non ne vuole sapere! E non è nemmeno il brano del caso che magari è meme o che; è proprio che la pura idea di questo fatto mi fa pisciare. Boh, o sarà il pacchetto emoji di Windows 10 che è particolarmente buffo a vedersi, o altrimenti ormai è ufficiale che anche il mio senso dell’umorismo, così come altri tratti della mia personalità, si è corrotto… ma ormai l’unica cosa importante è che non si corrompa l’archivio del mio Pignio!!! (E pure se succede, di quello ho frequenti backup.) 🤗

#Dev #devlog #Pignio




How to disable Firefox's battery-draining AI features


A few settings you may wish to consider for your firefox's about:config page.
browser.ml.chat.enabled = false  
browser.ml.chat.shortcuts = false  
browser.ml.chat.shortcuts.custom = false  
browser.ml.chat.sidebar = false  
browser.ml.enable = false  
browser.tabs.groups.smart.enabled = false  



Ecosia teaches how to activate ads for Ublock, but not for Ublock Origin.


Ecosia asks to disable ad blockers so they can earn money to plant trees, but I found that they explain it for several ad blockers, including Ublock, but not Ublock Origin. Do they think it's the same?


Karate or Tae Kwon Do for kids?


Hi everyone! So, my niece is alsmost 6 years old and is very energetic and generaly active as a child. Her parents are thinking of sending her to either Karate or Tae Kwon Do, both for physical and spiritual exercise and development.
Which one do you think could fit better for her age and also considering she likes it which is better in the long term?
in reply to WeAreAllOne

Karate is far better if teaching actual self defense is part of the goal. Tae Kwon Do is very questionable in terms of application outside the sport context. Of course, caveat is that, as with anything, it also very much depends on the skill of the instructor.


Is it possible to run qbittorrent and protonvpn in a VM?


Does anyone know how to run qbittorrent and protonvpn in a VM? When I try to run the qbittorrent setup app I get this message (image below) and I don't see anything mentioning a VM in the qbittorrent [dot] org forum.

I am new to torrenting, so I don't really know what to do. I figured/assumed that torrenting/seeding in a VM might be safer as it is another layer deep, and that it may help keep traffic separate (inside the VM: I'd be using a vpn and torrenting, and outside the VM: I'd not be using a vpn and just regular internet surfing). Is this possible?

Thank you.

in reply to Yourname942

Don't run your torrent client in a VM, that doesn't actually provide you with any additional security.

Use a Docker container instead. Binhex has torrent+vpn containers that will fetch the random open port number from Proton and pipe it into qBittorrent for you, as well as make sure the port is updated if the VPN drops. The container also acts as a killswitch.

in reply to _cryptagion [he/him]

Using a docker container provides you with the exact amount of extra protection as using a VM: zilch.

Only advantage is you can use other people's config easily.

  • signed, someone happily using their own VM-based setup


Study: Social media probably can’t be fixed


It's no secret that much of social media has become profoundly dysfunctional. Rather than bringing us together into one utopian public square and fostering a healthy exchange of ideas, these platforms too often create filter bubbles or echo chambers. A small number of high-profile users garner the lion's share of attention and influence, and the algorithms designed to maximize engagement end up merely amplifying outrage and conflict, ensuring the dominance of the loudest and most extreme users—thereby increasing polarization even more.

Numerous platform-level intervention strategies have been proposed to combat these issues, but according to a preprint posted to the physics arXiv, none of them are likely to be effective. And it's not the fault of much-hated algorithms, non-chronological feeds, or our human proclivity for seeking out negativity. Rather, the dynamics that give rise to all those negative outcomes are structurally embedded in the very architecture of social media. So we're probably doomed to endless toxic feedback loops unless someone hits upon a brilliant fundamental redesign that manages to change those dynamics.

Co-authors Petter Törnberg and Maik Larooij of the University of Amsterdam wanted to learn more about the mechanisms that give rise to the worst aspects of social media: the partisan echo chambers, the concentration of influence among a small group of elite users (attention inequality), and the amplification of the most extreme divisive voices. So they combined standard agent-based modeling with large language models (LLMs), essentially creating little AI personas to simulate online social media behavior. "What we found is that we didn't need to put any algorithms in, we didn't need to massage the model," Törnberg told Ars. "It just came out of the baseline model, all of these dynamics."