Study: Social media probably can’t be fixed "The [structural] mechanism producing these problematic outcomes is really robust and hard to resolve." arstechnica.com/science/2025/0…
they used llm bots to simulate that social media networks are prone to being the terrible mess that they are, even without algorithems. I would assume that these llms are built based on data largely of those terrible social media networks. It stands to reason that this simulation will result in modern social media, regardless of how you run the experiments.
Load an image with Exif orientation set to vertical using gdk_pixbuf_new_from_file(). (take a picture vertically with your phone or camera) With 2.42.12 the image is...
No creo, Signal no es tan buena como Conversations por ejemplo- O que cualquier xmpp/jabber. El que le ha puesto la info a la IA debe ser que no conoce conversations o similares.
Der Moment, wo die KI von Meta ein anderes Projekt empfiehlt, anstatt ihr eigenes. 😅 Ich persönlich würde allerdings an erster Stelle @simplex empfehlen und dann @signalapp.
The ability to switch between NVDA profiles with different keystrokes and to tweak output settings for each one individually is just fantastic. Whenever I mess something up on my main audio interface, I can get things working again in moments by switching to my "rescue" profile (that's just what I call it) which is hard routed to always use my laptop speakers. Highly recommend this for anyone who plays around with virtual mixers, audio cables, etc. It did save my ass more times than I can count!
I'm curious. Does this approach have different advantages than just setting NVDA to always use a static audio device? I have NVDA using a static audio device, and only switch it to the new one when I can confirm that the new one has audio working through it.
@BTyson For me it does. When I accidentally set the wrong virtual mixer in the Focusrite app, I lose audio because the device NVDA is using isn't being routed through my monitors / headphones. When that happens, the only way I can get speech back is by changing the mixer back inside the app. That's where the rescue profile comes into play.
It felt very weird to be on a plane that had a server with Nginx and Drupal installed on it. According to http headers at least, and those rarely lie. Definitely not a ground link, you aren't getting 5ms pings over satellite.
Their IFE system appears to serve MP3s ripped straight from iTunes (no DRM), with album metadata and such left intact. Is this legal? God only knows.
It's concerning to me that arguably *the* way to get Linux apps, Flathub, has all of its packaging data hosted on GitHub, with seemingly no plans to move away from it. With the direction GitHub is going, I am worried that Flathub will want to move and it'll be too late to do it cleanly.
To protect the digital foundation of essential government services, governments should invest in Open Source as public infrastructure and shift from consumption to contribution.
I fear that comes up against the buffers of present party politics.
Even situations like LocaGov Drupal that wouldn't exist witout Central Government funding. After they want it to "stand on it's own feet'.
Present political drift seems to be investment payments might be ok, but 'subsidy' (ongoing payments) bad. What is defined as (essential) infrastructure that needs subsidy is getting wittled away all the time.
it's a subsidy, though not usually called it. It's a subsidy as it is for rail infrastructure. Even the subsidies for roads, beloved of the right, now see: charging, tolls, privitization & generally trying to reduce direct gov costs. Even if less quick than it's been done to rail. Which is why I wondered about running up against the buffers of present politics. But maybe it's the rallying cry that will motivate people to change governments.
If I dig harder into the LocalGov Drupal example. The ongoing development is supported by local government - directly, through time and involvement; and indirectly through contracted agencies. But I think the success of that is because of quite some individuals, the way those organisations work together and not any concrete political direction.
To poslední, co od GPT asistenta chci, aby měl rozvernou „osobnost“, aby mě bavil jako kašpárek. Chci po něm informace, rychlé rešerše, pomoc při automatizaci činností, programování apod.
Jestli není u ostrých kritiků chyba spíše na straně klávesnice.
Je rozhodne znat, ze "mluvi" jinak. Ale to je spis proto, aby se trochu odlisil. AI styl psani byl podvedome tak rozpoznatelny, ze jsem na nej zacinal mit "AI slepotu" ruzne po internetech. Lidi, co se trefili do false-positive boxiku je mi lito.
There y'all go. Since a few asked for it, and I realized none of my code shared similarities in design or philosophy to the BTSpeak's tool, I released my Radio-browser Python CLI tool here: github.com/tgeczy/radio-browse… - useful on Raspberry Pi or other Linux distributions where just having a simple radio station player is what's missing. Can't say it'll get any mooer fancy than this, but we'll see! Took like 5 or 6 days to write it up in the end, wasn't just going to throw it all to AI to write without thorough supervision and fixing bugs, errors.
A Cli package for the public Radio browser API, built to be lightweight , accessible and easy to use from the ground up. - GitHub - tgeczy/radio-browser-whiptail-cli: A Cli package for the public ...
NV Access is pleased to announce that version 2025.2 of NVDA, the free screen reader for Microsoft Windows, is now available for download. We encourage all users to upgrade to this version.
This release includes improvements in Windows 11, browse mode, and Microsoft Word. Plus improved Braille display support, updated LibLouis and eSpeak-NG & more!
I decided to update on a portable coppy. I'm still uncertain about doing my main on this dell, as last time i updated, this coppy i have installed went bad, and speech no longer output. Requiring a family member to help me reinstall. It's not NVDA's fault, it's something to do with my sound driver. My sound driver is Wav Max audio, whitch is notorious for memory leaks. Portable coppies seem to do the best after updates, when it comes to this little computer.
@lexipic We've definitely seen those problems, althoguh I haven't seen any issues with wave max audio in a while - but I'm glad you got it behaving again!
my version of the driver is 5 years old. The computer itself was made back in 2017, and back then, the driver was Notorious for having memory leak issues, as well as audio device issues. I did attempt to disable it, and that has fixed the issue someone. However, that doesn't stop it from having a fun time every once in awhile. They did fix it recently, however, my computer cannot get the updates for the driver.
Here's a free tip, which I hope may help in future. It's odd, but works. I recently installed something that broke sound devices, so NVDA went to nowhere. Thing is, I *always* have it connected to a remote server, so I just opened up the non-official NVDA Remote app on iPhone, connected to my PC and let my phone do the talking. I didn't need to connect a keyboard to the phone of course, since I was still at the physical machine, but just used it to reset my sound devices and get things going again. If you set it up so that your machine connects with your custom key to a server of some kind, and you have either another computer, or your phone to-hand, you *should* be able to bring speech back without help in-future.
if you want to make AI worse, you can push shitty code on GitHub with all the data-training settings enabled. This is one most of us will get by (and will actually support you)
Do you have a pointer of what those early commit are? Currently looking into this from Forgejo's side. It's very possible that the counting function is missing a edge-case.
Having followed your posts about git, repo stats etc over the years, I'm sure you will be a positive boost highlighting things that can make and even better.
The use of AI is becoming common place at work, in school, and at home. The American Foundation for the Blind (AFB) is currently conducting the survey "Exploring Artificial Intelligence Use by People With and Without Disabilities." Complete the survey and share how you use AI: afb.org/research-and-initiativ…
A note of caution. If you begin to complete the survey, and select that you do not live in the United States, you receive a statement to say that you "Do not qualify for this particular study". As someone who uses AI extensively and daily, and who has almost delivered 12 hours of highly concentrated training on the subject of AI to blind people, I think I have a great deal to contribute. They could learn a great deal from me. But if they're not interested, then, so be it.
@brian_hartgen Sorry, but those regional restrictions are absolutely disgusting. Netflix and such services justify them with copyright laws, but what could justify it here?
@menelion Someone suggested it was about funding being restricted to US citizens. If that is really true, all they need to do is be upfront and honest; say so from the outset as an opening statement.
@menelion @brian_hartgen And I would say something that would reveal you in the midst of the commercial war. The states are not the only one who has AI. Rise up, China and the UE!
@luiscarlosgonzalez @menelion @brian_hartgen We understand the frustration that the survey not being made available to those outside the US. This is not a survey that we are conducting or have any input in, however, we thought it was interesting so we shared it. If you have feedback on using AI in work, school, at home, we would love to hear it. You can share it here or with our training team at Training@Vispero.com.
You can store a decade of email for a million people—call it 10-20Gb each—in a 10-20Pb storage array that costs under £1M per year (and is constantly getting cheaper). 10 year old emails are *cold* storage: drives not even spinning most of the time: you only really need 5-10% of it available on demand.
Nachdem es in den 2000er Jahren durchaus Handys von deutschen Firmen gab, ist es jahrelang sehr ruhig um Mobilgeräte aus Deutschland geworden. Umso spannende...
Just moved al my #git repos from #GitHub to my own #Forgejo instance. I did complete setup with #sso (single sign on) using #KeyCloak and with in-docker runner.
Up until today's morning I was going to install gitlab, but I was persuaded by being presented as lightweight, fully selfhostable, 100% open and with federating features on the way.
I'm also tempted to selfhost, but I think I'm going to stay with @Codeberg. For personal repos Forgejo is more than enough. I actually really like it: simple, clean and 100% open source.
@tosbourn very straightforward. I am running it using docker-compose, same as all my other selfhosted services. The only issue was smtp connection as I provided wrong PROTOCOL property value (starttls instead smtp+starttls). But even this was solved by a 10min web search. It has good config documentation. 👍
remember you only need a forge to share work with a larger team or open source. otherwise just push bare git repos via scp. one less thing to maintain.
BREAKING: we have just completely removed all ads from OSNews. The ad industry sucks, so I've put my money where my mouth is. OSNews is now entirely ad-free, for everyone.
Since this means losing a chunk of income, I would greatly appreciate it if you could support us by becoming a Patreon, by making an individual donation through Ko-Fi, or by buying our merch.
Accrescent is a novel Android app store focused on security, privacy, and usability. We’ve been building Accrescent since 2021 to improve the security and freedom of everyday Android users by providing an alternative to Google Play which doesn’t comp…
🌱 150+ people, 0 venture capital, 100% open source
In the recent interview with Techlore, Frank Karlitschek, founder of #Nextcloud, uncovers the evolution of Nextcloud, the challenges of open source businesses, and why digital sovereignty matters.
Mathijs
in reply to Ars Technica • • •