here's the deal @mozilla , the only people who care or even know about mozilla at all think this is bad to do. just say you fucked up, own it, and say everyone who uses firefox owes us $50 and we'll pretend it never happened. you took everyone who thought this was a good idea, fired them, and all the money will just be spent on making firefox work good forever and that's that. my fee is zero dollars and you can write that off your executive salaries i think because i am a lawyer too
the "Project 2025" thing, addressing reader
I'm sure that there are many on here that I don't need to tell this. But in case you haven't been following what's going on with the "Project 2025" thing:
It's a plan for a coup, to install a repressive dictatorship and police state in the US. Not metaphorically, not implicitly, but *literally*. Completely deconstructing all checks and balances and centralizing all government power into one person, Trump.
This is an actual, real threat - and a lot of the conservative crap lately ties directly into this plan.
So if you don't normally get very involved with politics, and have to choose only one political thing to care about, let it be this one. This is the fascist coup in the making.
And most importantly, you need to figure out at what point you're willing to use force and potentially violence to prevent this from happening. Figure it out fast.
I can't make that decision for you, but you need to have *some* answer to that question ready ahead of time. Don't leave it as "I'll figure it out when it comes to that", because then you'll be too late to prepare.
I am so grateful to @vantablack for being vocal and starting #fedipact. Opposing Facebook's control over social media always has been and always will be an act of asserting queer existence.
#FUSE question: are you supposed to ensure that the contents of any already-open file descriptors remain readable/writable even when the underlying file has nominally been removed from the filesystem? Or how is this supposed to work?
New blog post: Why picspam and repost accounts are bad! https://out.flipping.rocks/blog/why-picspam-and-repost-accounts-are-bad
> You've seen them on Twitter, you're seeing them on the fediverse: accounts with names like "AmazingNature8" or "HistoryPix", or occasionally human-sounding names like "Massimo" and "Veronique". Popular topics include nature and wildlife photos, historic images, relatable memes, fine art, and so on.
>
> They repost photos or memes that have gone viral elsewhere, like Twitter, Instagram, or Reddit. The text of the post is scraped as well; google a distinctive sentence and you'll turn up a previous post. Credit, if it's ever given, is a handle with no URL or indication of what social network it belongs to. Tellingly, on the fediverse, there are virtually never image descriptions. […]
>
> Since I end up having to give this spiel every time I raise the alarm about such accounts, I'm putting it down here in a more permanent place.
Are there any lawyers in #Ottawa on fedi that will witness me signing my name/gender change forms? So far nobody has responded to my emails.
Edit: thanks everyone, I may have a lead on someone who specifically works with a local LGBTQ+ group to get this done: https://kindspace.ca/idclinic/
practicing solidarity vs demanding unity
what I love about “practicing solidarity” as a frame is that it affords everyone agency, and it acknowledges that this is a set of skills and way of life that requires practice. it’s not a one and done hierarchical demand from on high. it’s a call to the circle of humanity and an earnest ask of each of us: look around, who needs the protection of the group, how can we best leverage our diversity on behalf of our most vulnerable
Looking for a post/paper I saw noting that the main difference between LLMs is their training data - same data, same outputs. (Because LLMs are lossy text compression.) Does anyone have the post/paper I'm thinking of?
EDIT: this is the one i was thinking of: https://nonint.com/2023/06/10/the-it-in-ai-models-is-the-dataset/
> model behavior is not determined by architecture, hyperparameters, or optimizer choices. It’s determined by your dataset, nothing else. Everything else is a means to an end in efficiently delivery compute to approximating that dataset.
LGBT and Marginalized Voices Are Not Welcome on Threads https://www.macstories.net/stories/lgbt-and-marginalized-voices-are-not-welcome-on-threads/
Here’s a story that I wish I didn’t have to write. But it’s high time that we make this a central point anytime, anywhere Threads is pictured as a decent alternative to Twitter and Mastodon. It is only that if you’re a straight, white male.
Mozilla, re: labour exploitation
A quick follow-up: digging into the history of CrowdFlower a bit, this exploitative data labelling company claimed Mozilla as one of its customers: https://web.archive.org/web/20170515113522/https://www.crowdflower.com/success-stories/
The same Mozilla that is now loudly talking about "ethical AI", and seems to have never spoken about their past dealings with this company. It's also not clear whether they still do business with them (it's part of Appen now).
Now it's not *certain* that this claim is true - tech companies, especially of the techbro kind, are certainly not beyond embellishing their customer lists to look more respectable. But this certainly raises some questions - especially since CrowdFlower legitimately has at least some large companies as its customer.
AI, labour exploitation, 'progressive' tech, history, long
So after watching the Ghost Worker documentary (https://www.youtube.com/watch?v=VPSZFUiElls) and finding that Lukas Biewald (featured in the documentary) seems to have written his own puffy Wikipedia article, I dug into the guy a bit more, and I ran across a thing he'd made in the past - the GiveWork app.
To quote from a picture (attached) explaining how that app worked: "Ever wonder if you could use a few spare minutes to do good for the world? [...] When you complete a task on your iPhone or iPod Touch, the same task is assigned to a marginalized person in Sub-Saharan Africa, South Asia, or Haiti. When your answers match and the task is verified, this person gets paid for the work you did together."
This app seems to have been widely covered at the time as an altruistic endeavour, a progressive use of tech, a way to "support refugees". It was developed and run by CrowdFlower, Biewald's for-profit company, and a 'non-profit' called Samasource.
Now you might have already noticed that that's kind of a weird model to be using to 'support refugees'. Why all the extra steps? (If you're familiar with the subject matter, you probably already know the answer by this point.)
Well, guess what Lukas Biewald, now the founder of Figure Eight (formerly CrowdFlower), was featured in the documentary for?
Quoting from an old conference recording shown in the documentary, Biewald speaking: "Before the internet, it would be really difficult to find someone, sit them down for 10 minutes and get them to work for you, and then fire them after those 10 minutes. But, with technology, you can actually find them, pay them a tiny amount of money, uhm, and then get rid of them when you don't need them anymore."
This makes it transparently obvious what the actual goal was of GiveWork: it was never meant to support refugees, it was meant to *exploit* them as a cheap source of labour. Each bit of data had to be processed by two independent people to spot wrong/fake data, and this was an easy way to both cut down on labour costs *and* frame it as a charitable act instead of the labour exploitation that it actually is.
But the reporting about GiveWork didn't mention that bit. Instead, it uncritically copied the framing of charitable, progressive tech.
Unsurprisingly, Biewald was featured in the Ghost Workers documentary because people doing work for his company where paid far below minimum wage, with Biewald trying to dodge any critical questions about that, explicitly only wanting to talk about AI.
Oh, and that non-profit Samasource?
"Samasource Impact Sourcing, Inc., formerly known as Samasource and Sama, is a training-data company, focusing on annotating data for artificial intelligence algorithms. [...] First founded as a non-profit in 2008, Sama adopted a hybrid business model in 2019, becoming a for-profit business with the previous non-profit organization becoming a shareholder."
---
And this is exactly the problem with 'liberal' approaches to progressive politics, that are not actually led by marginalized folks but by privileged folks who are out to 'save' them through a business, through tech, etc. (or at least claim so) - almost without exception, it turns into some exploitative bullshit, often deliberately so.
And since people rarely pay attention to what some organization does after the initial wave of "look what amazing progress they are making", there are essentially zero consequences or accountability. I bet that you didn't know about how GiveWork turned out, for example.
re: Mozilla, re: labour exploitation
(I originally found this information at http://faircrowd.work/platform/crowdflower/, which is apparently a crowdworker union, and verified it via the Wayback Machine)
Technical debt collector and general hype-hater. Early 30s, non-binary, ND, poly, relationship anarchist, generally queer.
- No alt text (request) = no boost.
- Boosts OK for all boostable posts.
- DMs are open.
- Flirting welcome, but be explicit if you want something out of it!
- The devil doesn't need an advocate; no combative arguing in my mentions.
Sometimes horny on main (behind CW), very much into kink (bondage, freeuse, CNC, and other stuff), and believe it or not, very much a submissive bottom :p
My spoons are limited, so I may not always have the energy to respond to messages.
Strong views about abolishing oppression, hierarchy, agency, and self-governance - but I also trust people by default and give them room to grow, unless they give me reason not to. That all also applies to technology and how it's built.