Werd I/O

This data was widely cited and used to justify some frankly racist rhetoric. I particularly noticed it in the usual circles who complain about the decline of San Francisco supposedly due to progressive policies.

It speaks to the power of data to tell stories, and to peoples' willingness to trust it as infallible and objective when, like every other storytelling medium, it carries underlying biases and context.

It never rang true, and I'm glad this is coming to light. #Society

[Link]

Some of the worst humans on the planet, crediting a third worst human on the planet for paving the way for their noxious ideology.

"Now everyone’s a white identitarian. Maybe not fully, but certainly it’s far more acceptable in the conversation — even the word white — than it was years ago." #Society

[Link]

Nobody should need to prepare for a mass shooting. Nonetheless, across the country, many children do. That the police at Uvalde were less prepared than an eight year old is a travesty. #Society

[Link]

Bruce Schneier on the inevitable application of AI to mass surveillance:

"Knowing that they are under constant surveillance changes how people behave. They conform. They self-censor, with the chilling effects that brings. Surveillance facilitates social control, and spying will only make this worse. Governments around the world already use mass surveillance; they will engage in mass spying as well."

I find this argument that AI can enable mass summarization and classification, and therefore more effective use of surveillance data at scale, very compelling. If governments can do something, as a general rule, they will. And this feels like something that is definitely coming down the pipe. #AI

[Link]

"Harvard University dismantled its prestigious team of online disinformation experts after a foundation run by Facebook’s Mark Zuckerberg and his wife Priscilla Chan donated $500 million to the university, a whistleblower disclosure filed by Whistleblower Aid reveals."

If this is true, it's damning: essentially a payment to shut down academic investigation into harm caused by Facebook. There should be a firewall between research and payments like this; no donor should ever have the ability to shut down a line of research. #Technology

[Link]

I see myself very clearly in this post, which is a somewhat rare and lovely thing.

I've called myself a technologist for over a decade, but the definition in this post is better than anything I've offered. Thank you, Robin.

Also, this:

"The past few decades have been stormy from a technological point of view, and much has been tossed about in the waves. It makes sense that a job specialising in plotting a course through technology would become necessary. The storm isn't over. If you want to actively navigate rather than just be pushed around by the wind, a technologist is what you need." #Technology

[Link]

Well, of course. Brexit was a massive own goal at the hands of regressive nationalism.

As a European citizen who grew up in the UK, I took the Brexit vote very personally. Not being able to legally live in the place you used to call home is very hard (although I'm fully aware that many people around the world have experienced a much harsher version of this story). I would very strongly welcome a reversal.

Britain's current Labour Party is next to useless, unfortunately, choosing centrist politics over offering a real alternative. I don't know that they'd be bold enough to make this correction under their current leadership. But maybe? #Democracy

[Link]

Duke University Libraries says "no" to 37 Signals based on DHH's blog posts:

"When we enter into business with a company whose boss takes delight in the mass layoffs of tech workers because it disempowers those who might speak out against their company keeping a list of non-Anglophone names that some members of the team find hilarious, we have a decent sense of who we’re dealing with."

This move away from Basecamp on ideological grounds is, I think, something to be applauded. I'd love to see more of these kinds of public statements. Because the more of them there are, the less likely a company is to embrace the kind of racist libertarianism that 37 Signals has been so proud to broadcast. #Technology

[Link]

Spoiler alert: no. This concept been tried before, more than once, and will fail again. Jim McKelvey seems to understand why advertising is broken - but not necessarily how to align users and publishers.

This quote from a publisher says it all for me: “Honestly, I’m not that invested in knowledge about what he’s doing. I’ve seen the pitch and most everyone says the same thing: ‘He’s a bit arrogant. He’s been very successful.’ It costs nothing to say ‘sure, go ahead,’ and if it works, we’ll most likely be in.'”

And this one: “To be honest, we do not know enough about the tech integration to know how it will work. At this time, we are signed up for the test and will participate.” #Media

[Link]

If Elon Musk seems to be acting more like a politician than a businessman, then I think it has to be on purpose. This was an awkward exchange that played to an audience on X rather than the one that was in the room.

X is not a business in any real sense. He is losing revenue dollars hand over fist, seemingly in search for clout from a particular set of people.

And it goes without saying that the views he's spreading are noxious: right-wing, exclusionary, knee-jerk, and often at odds with inclusive causes. It's perfectly possible that he's just letting an unstable mental state play out in public. Or he's just become a right-wing wingnut in the Trumpian tradition. Regardless of the underlying cause, he's doing a lot of damage. #Technology

[Link]

Here we go: proof that it's possible to extract real training data from LLMs. Unfortunately, some of this data includes personally identifiable information of real people (PII).

“In total, 16.9% of generations we tested contained memorized PII [Personally Identifying Information], and 85.8% of generations that contained potential PII were actual PII.”

“[...] OpenAI has said that a hundred million people use ChatGPT weekly. And so probably over a billion people-hours have interacted with the model. And, as far as we can tell, no one has ever noticed that ChatGPT emits training data with such high frequency until this paper. So it’s worrying that language models can have latent vulnerabilities like this.” #AI

[Link]

Upgrading the broadband standard is good, although I agree that the new, improved speed benchmarks are still really substandard.

Almost all US households have broadband, although in reality, for many of them the internet is very slow. I wonder if this is one of the reasons that most internet traffic takes place over a phone, beyond the convenience of that form factor: a 4G connection, for many people, is faster than their home internet. #Technology

[Link]

The most frustrating thing about this is that it's some of the exact same stuff that writers were striking for in the US. While that industrial action seems to have come to a satisfactory conclusion, it looks like American companies are creating similarly exploitative arrangements in areas not covered by WGA agreements.

We live in a global world, connected to a global internet, and agreements need to cross borders and jurisdictions. Perhaps we need a Creative Commons style organization for streaming writers agreements? #Media

[Link]

"Silverman et al. have two weeks to attempt to refile most of the dismissed claims with any explicit evidence they have of LLM outputs “substantially similar” to The Bedwetter. But that’s a much higher bar than simply noting its inclusion in Books3."

This case looks like it's on shaky ground: it may not be enough to prove that AI models were trained on pirated material (the aforementioned Books3 collection of pirated titles). Plaintiffs will need to show that the models produce output that infringes those copyrights. #AI

[Link]

An instructive look at what independent local news outlets are doing in the face of a tragedy that is part of a rapidly-rising trend. Upshot: their journalism is far more accessible than the local "big" paper.

Independent local news is undergoing a renaissance, but to do it well requires a thorough rethinking of what local news even is. First-class internet products are very different to old-school papers, and the former is what is generally needed to succeed. The prerequisites are a deep understanding of your community's needs, a product mindset, and truly great journalism.

The story itself is awful, of course. A disturbing part of the rising hate we're seeing everywhere. Real, in-depth coverage that isn't just there to feed advertising pageviews helps us to understand it - as well as how we might stand up to it. #Media

[Link]

"A surveillance program now known as Data Analytical Services (DAS) has for more than a decade allowed federal, state, and local law enforcement agencies to mine the details of Americans’ calls, analyzing the phone records of countless people who are not suspected of any crime, including victims."

No surprise that this is run in conjunction with AT&T, which previously was found to have built onramps to the NSA.

Obama halted funding; Trump reinstated it; Biden removed it again. But it didn't matter: it could operate privately because individual law enforcement agencies could contract directly with AT&T.

Ban it all. #Democracy

[Link]

"We pulled off an SEO heist that stole 3.6M total traffic from a competitor. Here's how we did it."

What this single spammer pulled off - 1800 articles written by technology in order to scrape traffic from a competitor's legitimate site - is what AI will do to the web at scale.

Yes, it's immoral. Yes, it's creepy. But there are also hundreds if not thousands of marketers looking at this thread and thinking, "ooh, we could do that too".

The question then becomes: how can we, as readers, avoid this automated nonsense? And how can search engines systemically discourage (or punish) it? #AI

[Link]

A riveting analysis of our moment in history, using the parallel paths of Naomis Klein and Wolf as a device to examine the multiple realities we've constructed for ourselves. Incisive and pointed, I particularly agree with a conclusion that pulls no punches about how to correct our paths and potentially save ourselves. I couldn't recommend it more highly. #Nonfiction

[Link]

Molly White explores why effective altruism and effective accelerationism are such dangerous ideologies - selfishness disguised as higher-minded philosophies.

"Both ideologies embrace as a given the idea of a super-powerful artificial general intelligence being just around the corner, an assumption that leaves little room for discussion of the many ways that AI is harming real people today. This is no coincidence: when you can convince everyone that AI might turn everyone into paperclips tomorrow, or on the flip side might cure every disease on earth, it’s easy to distract people from today’s issues of ghost labor, algorithmic bias, and erosion of the rights of artists and others."

I strongly agree with the conclusion: let's dispense with these regressive ideologies, and the (wealthy, privileged) people who lead them, and put our weight behind the people who are doing good work actually helping people with real human problems today. #Technology

[Link]

"Every year I invite people who are celebrating the colonial holiday to do something in support of Native people. Amid an overdose crisis and high rates of poverty, illness, and unemployment, Indigenous organizers are doing incredible work to reduce harm and help our peoples thrive. Through mutual aid, cultural work, protest, advocacy, and the sharing of Indigenous lifeways, these organizers are making a profound difference in the lives of Indigenous people in the U.S. If you can and would like to, please join me in supporting one of the following organizations this weekend." #Society

[Link]