A million people have talked about it, but something very grim and interconnected about the ever-growing cis/right-wing obsession with, well, gender. More gender for them, less for the rest of us.

++ a great reference on some of the racist pseudoscience cited in these summaries.

bsky.app/profile/stai...

Awesome article and excellent reference, thank you for sharing. Grok is definitely worse, probably due to tone prompting Elon's team has built in. But it strikes me that the corpus they're drawing from is similar -- even the Google answer says that heritability of racial IQ differences is "high."

Anyway, I'll plug my own post, where I wrote about similar AI search problems and anti-trans misinfo.

bsky.app/profile/beee...

The tendency of "censorship" to dredge up dumb websites made me think of this article by @evanup.bsky.social + colleagues on "dredge words" which, when entered into search engines, dredge up stupid and harmful websites.

arxiv.org/abs/2406.11423

Relatedly, prior research has shown that partisan search terms will surface partisan results, potentially further entrenching people in their misguided views.

Here's one Google AI Mode answer to "hbd censorship," which cites a race science Substack and a far-right manifesto as its sources.

This is probably not due to any improvement in LLM technology, and probably due instead to these companies tinkering with their algorithms after critical coverage like this from @wired.com

www.wired.com/story/google...

Admittedly, most chatbots -- including desktop Grok! -- give (slightly) better answers to questions like these, or refuse to answer them at all.

If you Google questions related to race and IQ, you often won't get an answer. Most chatbots, like ChatGPT, do (relatively) better with this query.

The Meta answer cites a web page generated by Ithy, an LLM chatbot which supposedly combines results from other chatbots.

The TikTok video cites a Jordan Peterson clip.

The Google answer cites a paper by a former head of the white supremacist / eugenicist organization the Pioneer Fund.

Also, if you're unfamiliar with the disinformation campaign suggesting that the rate at which trans people detransition is very high, you can read this article by @juliaserano.bsky.social!

Naturally, not cited in any search summary I've seen yet...

juliaserano.medium.com/spotting-ant...

In this case, it probably picks up (false) claims that detransition research is flawed from sources associated with anti-transgender disinformation campaigns, also cited in the search summary.

Perhaps it then searches for abstract "limitations," even when they're not related to the topic. 🤷‍♀️

anyway i can't (can) believe these apps got away with providing us an interface to rate the traits of other human beings on a mass scale

Have learned loads from Jaylexia's work on gendered racial capitalism, and think about it a lot in my own work on far-right media. Similar to how gig work platforms profit from satisfying white users' racialized emotions, far-right media platforms profit from these emotions in white readers.

died reading this