On reviewing privacy preserving tools:
This is not a new discipline.
We have mathematical and engineering tools to do analysis.
We have decades on decades of research literature, rooted in cryptographic analysis, statistical methods, probability theory, and computer science detailing how privacy preserving system are broken.
Just how one can tell that a badly engineered bridge will collapse before it is built, one can assess that a "privacy preserving" tool will not preserve privacy.
Today seems like a really good day to remind you to get @signalapp
There's a funny thing you see in many scientific papers - especially #AI papers: The paper will prominently include a link to a GitHub repository with claims of code availability "soon" but when you go there (months after the paper was released) there's either just a placeholder or the paper text.
People use GitHub links to score browny points for "doing open science" but most of it is just not there. Especially with statistical systems when you realize that you don't get the training data, you don't get the code, you don't get model weights what you get are results and a "trust me bro".
@tante
An exception that proves the rule: https://www.llm360.ai/
They release everything you mentioned, plus intermediate checkpoints *mapped to the training data*, and various metrics.
Building a culture of accessibility - Felicity Miners-Jones @ Tetralogical:
https://tetralogical.com/blog/2024/04/19/building-a-culture-of-accessibility/
9 signs your frontend code has quality issues that affect your users - Angelika Tyborska:
https://angelika.me/2024/04/13/9-signs-your-frontend-code-has-quality-issues/
How to think about HTML responsive images - Dan Cătălin Burzo:
https://danburzo.ro/responsive-images-html/
Beware – automatic tools over-report accessibility issues and steal your time - Bogdan Cerovac:
https://cerovac.com/a11y/2024/04/beware-automatic-tools-over-report-accessibility-issues-and-steal-your-time/
uspol
@chjara
Kind of. The four bills separately passed in the House, and are being bundled for a vote in the Senate.
About 14 years ago, one of my coworkers read California labor law and realized we were entitled to overtime for working more than eight hours a day, rather than for 40 hours a week. It meant a big bump to my income, to about $30/hour. I was working night shift, with twelve hour shifts, and two hours of commuting each way. I was continuously sleep deprived and barely saw my family that I lived with, for five years.
@opensuse Factory became bit-by-bit reproducible, enhancing #Tumbleweed's verification. Thanks to all involved! 🚀 95% previously passed. @reprobuilds https://news.opensuse.org/2024/04/18/factory-bit-reproducible-builds/
Basic things which are: irrelevant while the project is small, a productivity multiplier when the project is large, and much harder to introduce down the line
https://matklad.github.io/2024/03/22/basic-things.html
Discussions: https://discu.eu/q/https://matklad.github.io/2024/03/22/basic-things.html
A spy tool is scraping the messages of thousands of Discord servers and selling the data. This is letting people track users across servers, shows when they joined voice chats, which servers they're in, etc: https://www.404media.co/a-spy-site-is-scraping-discord-and-selling-users-messages/
Wow, Marginalia's new article on Query Parsing and Understanding comes right as I'm about to start work on coding more query parsing features into Clew. Good timing.
The true power of #genAI is not technological, but rhetorical: almost all conversations about it are about what executives are saying it will do "one day" or "soon" rather than what we actually see (and of course no mention of business model which doesn't exist).
We are told to simultaneously believe AI is so "early days" as to excuse any lack of real usefulness, and that it is so established - even "too big to fail" - that we are not permitted to imagine a future without it.
Today I read https://www.scattered-thoughts.net/writing/pain-we-forgot/
and thought up several things I want from an IDE
For a pure language like Haskell, I want the IDE to magically find properties for me.
There are already libraries like https://github.com/nick8325/quickspec and https://github.com/rudymatela/speculate that can do it, but they're not yet hooked into haskell-language-server.
I also want my IDE to discover which tests execute which functions, and run the tests whenever the functions change.
Working with long scientific data pipelines can be a nightmare because the code of dozens of people, package managers (conda, pip, poetry) and workflow tools (make, snakemake, nextflow, pydoit) just bundle up into an indecipherable mess.
I'm wondering if people here know of good examples or have suggestions on how to keep scientific software workflows lean... Or at least make them less of a pain.
Increasingly finding that #AI-generated images on a blog post lowers my opinion of the author — not (just) because of a principled anti-AI stance but also simply because it’s rubbish.
If you’re happy decorating your work with generated garbage I’m just going to assume your standards are that low for everything you do.
@benjamineskola
If an article's text is good, but the images are AI-generated, I've found this is a warning sign for the text being outright plagiarism.
Tiny Privacy Tip for Application Developers 🔒✨
Every piece of data you
collect on others with your application becomes a liability to you.
You are responsible for
safeguarding and keeping track of every single piece of personal data you collect.
This is a heavy responsibility.
Especially if you collect and store a lot of data.
A much easier approach is to collect only what is absolutely necessary and delete it thoroughly as soon as it is not necessary to keep it anymore. You will save yourself so many headaches adopting this practice right from the start in your software development.
Remember: You can't be liable for the data you simply never had.
This is the easiest path for you,
and the safest path for your users.
en: Mostly tech, but not entirely. Privacy is a human right.
ia: Principalmente technologia, ma non in toto. Privacitate es un derecto human.