Show newer

re: poisoning AI by disregarding disabled people 

@Yza @clarfonthey I do not find that a credible idea. LLM companies are certainly aware of accessibility content, and that it is a good way to find relatively high-quality training data. They aren't just taking whatever content they come across without quality assessment.

And if you're relying on scraper traps anyway (assuming they work), then you don't need to pollute the accessibility information to begin with, and it is no longer relevant to the original post?

I don't think it's helpful to suggest vague solutions without concrete details on how it would work and avoid disaster, personally. It creates the appearance of there being solutions without actually solving any of the hard problems.

re: poisoning AI by disregarding disabled people 

@Yza @clarfonthey How would that even work, though? What stops an AI scraper from simply acting like an accessibility tool in its interpretation of the page?

poisoning AI by disregarding disabled people 

just posting this on its own since it probably got lost already, as a comment on a different thread:

basically every method of "poisoning" AIs by messing with content in "unnoticeable" ways for all clients, will make your content less accessible.

trying to analyse network traffic and put AI scrapers into mazes is fine and won't affect most people. adding hidden text "off screen" has a very high chance of affecting screen readers and other accessibility aids, while not affecting AI scrapers.

one key example is where people have proposed to modify subtitles in videos to add lots of garbage that is "invisible" or "off screen", similar to the age-old trick of having white text in the background of a PDF that gets picked up by résumé scanners. the solution that all of these tools have chosen is to just render the document and OCR the text, meaning that the bad actors continue as usual, but the good actors get messed up as their screen reader lists out all sorts of nonsense, and everyone suffers as the scanners fail to properly OCR text that isn't perfect Times New Roman Eleven Point Font.

for videos, people have argued that blind people can listen to the unmodified audio track, whereas deaf people can read the subtitles which are only showing what's on the screen. this excludes deaf-blind people who are forced to read the transcript with a braille reader: they will get the nonsense you put in the subtitles. or, just anyone who wants to read the transcript for any other reason.

it really is admirable for people to try and "poison" AI data, but unless you do so in a way that is tailored to the clients doing the scraping (network traffic), you are going to just fuck over people who need your accessibility data. please don't

re: transphobia 

@oddtail A lot, in my experience, especially privileged folks :(

I've lost count of how often I've had conversations where I pointed out that they have never apologized or even acknowledged wrongdoing, and especially privileged folks would accuse me of having unreasonable expectations because "they've clearly grown as a person"

transphobia 

You know what I "love"?

When people who were previously openly bigoted towards me feel a shift in attitudes and start appealing to my humanity or sense of community or empathy or good political sense without owning up to, or even acknowledging, their past wrongdoing.

Pretending to be the "bigger man" when no longer allowed to be shitty has got to be amongst the most disgusting tactics out there.

LLMs, article link, someone's personal vent that contains slurs 

Sharing this in the hope that it can help to get the point through to some LLM fanatics: ludic.mataroa.blog/blog/i-will

Note that it contains slurs that are ableist or antisemitic in origin, although I do not think that was the intention behind their use here.

there really needs to be a kind of nix that can run reasonably well on lower-power devices like the librem 5. plasma mobile is barely able to keep running while the nix config is being eval'd (and no, this isn't even the build yet, the eval part alone takes nearly 9 minutes and is enough to make this device struggle)

Ik wil graag een CW als er een foto van de oranje miscreant wordt gepost. Daar krijg ik namelijk stress van.

The first step of creating a #wheelchair model that isn't a total disgrace is to not forget about the human.

Step zero might be having friends in wheelchairs or having seen actual wheelchairs.

(if you want to know why this trot seems a bit salty, look for wheelchairs on sketchfab ...)

Feel free to give feedback if you use wheelchairs.
#3dart #blender #modeling

computer, linux help, :boost_requested: 

hi, I have a "server" that's occasionally completely freezing in correlation with LVM snapshots

Update: 2025-03-07 update has some more actually useful info, including the fact that lvremove on a snapshot followed by a freeze. and it somehow happened when i did a manual trigger backup

note: i have much more up to date information on my wiki. short link is w.jc6.xyz/9 because the long link is wiki.jacksonchen666.com/wiki/R :p

please check my wiki for complete information. below is kind of a summary version that's not complete.

i make LVM snapshots on backups, then remove them after it's done. it seems the last syslogs i usually get is a log that the backup is starting, then logs stop there. interestingly, a backup at the time is made and appears to be complete. after i restart the system, the next backup will be empty, then the second next backup will be normal (my script removes snapshots, and doesn't error on existing snapshot name when creating a snapshot)

i don't have logs of what my backups do, unfortunately. currently system logs are sent via network including kernel and syslogs

when it freezes, SSH is impossible and pressing keys on a keyboard does nothing (except for tty 7-8, which might use a different tty program), but the tty cursor still blinks, and system still responds to sysrq. but sysrq actions which dumps info doesn't really dump any info, and I can only reboot from there.

the system also seems to be idle when it freezes, given it's slightly colder?

running alpine linux 3.21.2, kernel 6.12.9-0-lts, lvm 2.03.29-r1 (distro package) (also all might be slightly outdated by now), feel free to ask any info I might've missed

saying “LLMs have their uses” is like saying “Agent Orange has its uses” - technically correct. more correct than LLMs often are, even, but folks, I hate to break it to you: “technically correct” is not actually the best kind of correct except in a very very narrow set of circumstances that you’ll rarely find yourself in. “technically correct” is just another version of the Torment Nexus.

I wrote a CDC Archive Archiver to pull public information from the Centers for Disease Control and Prevention website. Just in case people get weird ideas of removing science from the public discourse. This is part of a larger effort that, among others, @lavaeolus initiated. You are welcome to chime in if that's up your alley.

- codeberg.org/antifascistDH/aca
- fedihum.org/@lavaeolus/1138822
- codeberg.org/thgie/antifascist

#AntifascistArchives #DigitalHumanities #ArchivalPractices

Nieuwsartikel dat Buienradar en co het KNMI aanklagen omdat ze een 'te goede' app hebben gemaakt. Dat vinden ze concurrentievervalsing.

Ik wist helemaal niet dat het KNMI een goede weerapp had, maar nu wel, dus nu gebruik ik die i.p.v. Buienradar. Bedankt, Buienradar!

Something I've been noticing lately, in a lot of different places, is that as economic conditions get worse, things start getting stolen more often.

I'm not talking about shoplifting here; I'm talking about people stealing from their own communities. I've noticed increased reports of this in a wide variety of communities, across countries.

Please be mindful of who you are stealing from, and don't let it be your community. And conversely, if you are still doing well financially, look for people in your community who are financially drowning and how you can help them.

actionable, AI 

Massive tech companies have invested absurd amounts of money into "AI" systems, recentering their whole business around it.

It seems to me that one of the most effective ways to hurt these tech companies, would be to sabotage their play by making "AI" as socially unacceptable as possible, so that their investments turn into useless bricks that nobody wants or trusts.

: I have a Rust question. My usual approach to developing a collection of modular (but related) libraries in JS is:

1. Start with one monolithic package, in which everything is developed initially.
2. But opportunistically minimize assumptions between parts of the package, so that internally it is already mostly modular.
3. Spend some time figuring out reasonable APIs for the remaining, more-difficult-to-modularize cases.
4. Eventually split out all the internal parts into separate libraries/packages and publish them independently, so that users can select only the parts that they need, and easily combine it with third-party libraries.

What would a good workflow for this look like in Rust, especially in terms of internal crate layout (in step 1) so that it's easy to split out into separate crates later, but also in terms of tools to use?

I do not use monorepos, each crate will have its own repository in the end. Also, contributors must be able to use standard Cargo commands to manage the different crates, so no 'wrapper scripts' or third-party tools (unless they can be made to run from within Cargo).

:boost_requested:

Wow, a single comment questioning @dansup@mastodon.social and I was immediately blocked, not even a discussion man? really?

their choice to keep calling loops "open source" while the code is closed, as well as the recent post about hiding federation features behind extra menus so all new users auto-signup to services controlled by him, is concerning for genuine reasons (although I'm sure he's already called me a troll or something)

This combined with the comments/jokes about "exploring" advertising for the loops platform is highly concerning, here you have somebody siphoning users into a single server they control, while developing an app they call open source without any code being available, who also thinks advertisements are a reasonable business model for the fediverse...

We could have just spoken about your views on this, but I suppose my ideas about FOSS and no advertising were too threatening lol

Think twice, or thrice, about actually signing up and supporting
#Pixelfed or #Loops while dansup is behind them tbh

uspol 

With “DEIA” being dismantled, it’s time to rethink what we want. What we were doing wasn’t very effective; much of it put lipstick on a hierarchical pig. It’s time to think instead about what actually enables the thriving of the people targeted by white supremacy and what undermines the power of white supremacy.

To me, the hierarchical nature of most workplaces—a nature DEIA efforts worked within rather than challenged—is a big part of the problem. The bias of others who rate you and hand opportunities to you matters much more than the bias of a colleague you can associate with, or not, as you choose.

Another problem is the need to do something you hate doing because it helps nothing but an owner’s wallet. Owners *need* to control you, and an authoritarian power structure and infighting helps to achieve that control.

To achieve a world that is just and equitable and accessible, we need to radically rethink power relations, not establish yet another powerless committee.

Show older
Pixietown

Small server part of the pixie.town infrastructure. Registration is closed.