Link Roundup: March 2025

Revisiting Radway’s Reading the Romance: A Critical Ethnography of Romance Fans | Osteophage – Coyote discusses a 1980s book that studied women who read romance novels.  The author Janice Radway is sympathetic to the women of her study, seeing them as mistreated housewives trying to find an escape.  And yet, the romance books contain a lot of sexism themselves, and she is disappointed to find that the women tend to uncritically accept that sexism.  Coyote positions Radway’s book in relation to more recent debates about fandom.

Trans People are Under Attack and We Must Help Them | Rebecca Watson (video and transcript, 9 min) – Trans and nonbinary people are a tiny minority (estimated at 1.6% in the video), so how much does it matter that they’re under attack?  Well, that’s a lot of people if you think about it.  For instance, it’s far larger than the number of federal employees fired or laid off, and it’s larger than the total number of federal employees period.  Trump has signed 80-some executive orders, and if each one chips away at the rights of as many people, that affects all of us.  (And the video doesn’t even discuss the ways that cutting trans rights directly impacts cis women, e.g. by requiring them to undergo invasive examinations for sports.)

[Read more…]

Link Roundup: February 2025

Content note: I don’t have any links about the Trump administration today.

AIs Regurgitate Training Data | Reprobate Spreadsheet – Last month I wrote about the claim that AI “regurgitates training data”, and some people claim that this virtually never happens, or else they claim that it’s the only possible thing that happens.  And I keep saying, you don’t know either way!  It’s a question that can only be answered through empirical research.  And what the empirical research says, is that models do it sometimes–and that’s bad enough.  HJ discusses some of the research here.

But I have a bit of a critique.  HJ describes a study that asked an LLM to predict number sequences, such as currency exchange rates.  The predictions had lower root mean square error when predicting sequences in the training data.  The researchers call this “memorization”, and HJ calls it “regurgitation”, but I call it a textbook description of “overfitting”.  Clearly the models are retaining excessive unwanted information from their training sets, but calling it “memorization” creates a false impression that it’s verbatim quoting, which it’s not.

This is Arousal | No Pun Included (video, 20 min) – A board game critic traces a popular claim: the most fun part of a board game is opening the box, and then they read the rulebook where fun goes to die.  It’s based on a small study of families playing Hasbro, which measured physiological arousal rather than fun.  It’s not a strong study, but you know, it’s just a grad student’s proof of concept, it’s fine, been there.  It’s just wildly inappropriate to generalize into a nugget of conventional wisdom.  This video is a great example of science popularization done well in an unusual domain.

[Read more…]

Link Roundup: January 2025

This month, I wrote a history of color symbology in ace and aro flags.

Atheist group faces backlash after publishing, then removing, anti-trans article | Friendly Atheist (via) – If you’re interested to know how atheist organizations deal with trans issues, Hemant provides a pretty good summary, through the lens of one recent incident.  Some orgs are better than others, but it’s frustrating how even trusted orgs can’t seem to maintain a consistent trans-positive stance.  Thankfully there were enough trans-positive voices to pressure FFRF to retract.  But I think atheist orgs have likely suffered from evaporation, with many trans-positive folks simply opting to leave (like me!).

Matthew S. Burns on AI, Empathy, and the Making of Eliza | Circuits & Synapses – As people become broadly familiar with AI chatbots, most old fiction about AI has not aged very well.  After all, the fiction was never about AI, it was about exploring humanity.  Eliza is a visual novel about using AI for talk therapy, and I think is one of the few works of fiction that has aged very well, because it was grounded in the realities (and pitfalls) of tech.  This is an interview with the author, looking back at Eliza.

[Read more…]

Link Roundup: December 2024

This month, the ace journal club discussed genital plethysmography, and then we took seriously the idea of a sex therapy as a BDSM scene.

The Really Dark Truth About Bots | Benn Jordan (video, 29 min) – How much activity on X is from bots?  Benn Jordan goes through the research, explains how bots work, and concludes that it’s a whopping 1/3.  Incredible.  Kill it with fire.

I often think about the Rationalist/EA crowd, and how for many years they have been concerned about AI causing a human extinction event.  I don’t think the concern was entirely misplaced, but they seemed most concerned with the “AI alignment problem”, i.e. making sure AI does what we actually want instead of deciding to kill all humans.  However, I’m far more concerned about the billionaire alignment problem, i.e. making sure billionaires do what we actually want instead of deciding to kill all humans (now with AI assistance).  Or for that matter, the foreign dictator alignment problem.  All the AI safeguards in the world won’t help if powerful people simply don’t want them.

A comprehensive pro-choice ethic | Tell Me Why the World is Weird – Some pro-lifers advocate a “comprehensive pro-life ethic” where they advocate for the health of people already born.  You can think whatever you like about that, but Perfect Number turns it around and imagines what it would mean to have a comprehensive pro-choice ethic.  It would be about empowering people to make free and informed decisions on medical treatment and reproductive health.

[Read more…]

Link Roundup: November 2024

This link roundup does not include any discussion of the US elections, and I do not have plans to write about it in the immediate future.  My thought about it is: pace yourself!  We’re on a slow motion train wreck, don’t burn yourself out on the first week.

This month, I reviewed I Want to be a Wall–that’s the silly graphic novel I referred to earlier.

Cohost September 2024 Financial Debate Retrospective: Making Sense of The End | osteophage – The ad-free social media platform Cohost recently financially collapsed.  Why?  Coyote explains why many of the popular theories are incorrect.  Cohost was able to generate healthy revenue for its size, but its dev team had unrealistic expectations, trying to support four full time tech salaries.  Also the devs were trying to make a competitor to Patreon, but this is a doomed venture because it requires a great deal of regulatory compliance overhead that the devs weren’t even aware of.

Yeah, that just sounds like ordinary tech startup incompetence.  There’s nothing fundamentally impossible about what they were trying to do!  Other ad-free social networks exist.

The Visualizer’s Fallacy | Christian Scholz – After writing my post about Wittgenstein, I found someone who wrote a dissertation on Wittgenstein and aphantasia.  He observes that aphantasics can in fact think without visualizing, and they even perform well on shape rotation tests.  So does that mean visualization is unnecessary for mental rotation?

[Read more…]

Link Roundup: October 2024

How I Fell Out Of Love With Facebook | Tantacrul (video, 3:10 hours) – A comprehensive review of all the scandals that Facebook got involved in.  I had known about a few of these but hadn’t heard of the more international scandals, like Free Basics.  I mostly remember how Facebook sought to reduce bias on their platform, and through the funhouse mirror of corporate priorities it turned into refusing to take down politically conservative posts less they generate an appearance of bias.  Facebook really is a nightmare of corporate immorality.

Fast Crimes at Lambda School | Sandofsky – A long article about the scandals surrounding Lambda School, a coding boot camp.

My husband went to a bootcamp (under an income share agreement), and I went to something like a bootcamp (under a hiring fee model).  Both of us owe are career success to them.  There’s nothing about the idea that makes it inherently bad or unworkable.  But… the one my husband went to was exaggerating its job placement rates by excluding people they kicked out of the program, and excluding people who remained at their current job (!?).  The program I went through was more honest–but it all but collapsed during the pandemic.  Both of our programs were extremely selective.  From what I can tell, bootcamps operate on very thin margins, and are not easy to scale up.  It sounds like Lambda School immediately tried to scale up, and simply could not get its unit economics working, no matter how much they fleeced students.

[Read more…]

Link Roundup: September 2024

Did you know that some of my links, I get from other link roundups?  I don’t always give credit to aggregators, since it’s not like they wrote the original article.  But in case you like more links, the ones I follow are Critical Distance (games criticism), Perfect Number (ace ex-evangelical blogger), and Ozy (rationalist blogger).  I also run a separate link roundup for The Asexual Agenda.

Oh, and in case anyone is interested, I wrote a couple queer fiction book reviews this month: Aces Wild, and The Bell in the Fog.  On to the links:

Can You Trust An AI Press Release? | Asterisk Magazine – When I wrote about LLM error rates, I pointed out even when AI companies boast of their models’ performance, the error rates are there in plain sight.  But I also said you shouldn’t actually trust those numbers.  This article goes into more depth, explaining how AI companies can select information that shows their products in the best light, while understating the performance of rivals.

The Games Behind Your Government’s Next War | People Make Games (video, 1:12 hours) – A look at the world of wargaming, i.e. games made for the serious purpose of helping decision-makers prepare for war and other crises.  The video forthrightly confronts the ethical question: is this killing people?

I think my stance is fairly favorable to wargaming.  Assuming that wargaming is effective (although this is legitimately in question), I would really rather that decision-makers are good at strategy, rather than bad at it.  I wouldn’t celebrate an incompetent soldier for saving the lives of rival soldiers.  I think it’s a mistake to blame only the military for bad wars, when a lot of the blame belongs to the cultural and political systems that decide to make war in the first place.  The part I blame on the military is when they produce propaganda that tilts cultural sentiment to their own benefit.

[Read more…]