there are no “theory neutral” models and if your modeling paper paper starts with that as an “advantage” of your approach you have immediately lost me

Quite rare to see a chart that says quite so overtly that no one involved has the slightest clue what’s happening here

40,000-year-old proto-writing in Germany www.pnas.org/doi/10.1073/...

lol

replace connectionism with LLMs and you’re up to date

"These are [...] people who think they can go from the definition of conditional probability, via Harry Potter fanfic, to prophesying that an AI god will judge the quick and the dead, and condemn those who hindered the coming of the Last Day to the everlasting simulated-but-still-painful fire." 🔥

We asked people who lived in homeless encampments that were cleared out in city “sweeps” to write about what object was the hardest for them to lose. “They took my baby pictures and my moms obituaries,” a man in California wrote. (Published Dec. 2024)

hot take: the transformer is an inelegant architecture

It's a Motte & Bailey strategy. Their broad desire is the elimination of trans people from public life, and they've decided the easiest issues to toxify are sports & youths. The consequences have already expanded beyond that, and poison the discourse about the veracity of trans people.

threw my mfing phone across the room

I just got my annual Public Lending Right cheque for Because Internet from the Canada Council for the Arts, so permit me a thread about what the public lending right is and why you should request books you love from public libraries!

look what came in the mail

what is the point of stealing cuisines from the entire world if one of them is not maple syrup

linguistics

Bizarre to read an article that is ostensibly about the increase number of heart attacks in people in their 30s and 40s that has more to say about an aging population than it does repeat infections with a SARS virus known to damage the vascular system. Cool.

neat stuff; getting closer to figuring out why the heck we yawn... www.biorxiv.org/content/10.6...

mine = my + noun. This is how we know the linguistic capacity isn't encapsulated in the Fodorian sense because how else would the grammar know "noun" begins with an "n"?

People who work in tech and have been interviewing candidates; how explicitly are you looking into AI coding assistant usage/knowledge *at interview time* (i.e. do you expect it during your coding interview? do you ignore it entirely?)

these are amazing

the urge to stick a QR code that leads to this post on every ring doorbell in my complex

probably many or most people who follow me have read or skimmed this, but I was just prompted to go through it again and I still think it's a pretty sound stance... henryfarrell.net/large-ai-mod...

It's funny to me how a lot of folks have been insisting on "skeet" for nearly three years when "bleet" (like bleat) is right there. #UselessReflections

Foucault tried to warn us and people keep installing ring cameras it sucks

This is gonna be so cool (no way I can make the submission deadline and I live to far to just go but y all should!!!)

dry mouth january

I know many Normies who have seen someone, usually a guy, blow through thousands of dollars on sports betting Very much the origin story of a 21st-century temperance movement

ok so does the "glazing" slang come from where I think it does??!

i think probably we should be teaching and demonstrating cold reading to the ai people

the only Claude that matters to me

this bot social network is dumb and it's gonna make ppl say (more) dumb things

leaving aside its up-its-own-assedness the name is just too close to "moldbug" for comfort

A new study from Anthropic finds that gains in coding efficiency when relying on AI assistance did did not meet statistical significance; AI use noticeably degraded programmers’ understanding of what they were doing. Incredible.

Linguistics.

nightmare shit jfc

So initial experiments with Open AI's vibe-coding science tool Prism are going about as well as expected.

claudehopper

"The secret fear of the morally depraved is that virtue is actually common, and that they’re the ones who are alone."

counterfactuals are an example of the autonomy of syntax in the same way that "colorless [sic] green ideas" is (yes I know that wasn't the original point of that)

"pourquoi redire la même chose que tous les autres à des gens qui le savent déjà?" je sais pas mec, c'est une activité sociale normale entre humains? ça rassure et ça lie le groupe? ça aide à se sentir moins seul.e?

I wrote up some thoughts on why the Poverty of the Stimulus Argument - a mainstay in linguistics and other areas of CogSci - is not (really) an argument. I also argue, controversially, that usage-based reliance on LLMs to refute the POSA are weakened. 🧵 vincentcarchidi.substack.com/p/on-misunde...

lotta ppl doing Commander Keen dirty in the replies here

unsubscribe

I just created a series of seven deep-dive videos about AI, which I've posted to youtube and now here. 😊 Targeted to laypeople, they explore how LLMs work, what they can do, and what impacts they have on learning, well-being, disinformation, the workplace, the economy, and the environment.

wake up babe new "he was a punk she did ballet" just dropped

fascinating watching my company struggle to roll out a complex agentic system and slowly come to the realization that the unavoidable latency, cost, and non-determinism are anathema to both our own business model and that of the clients for our core product

AI prompts as a slot machine is a powerful observation. So much of this mania makes sene when you realise these people are addicted to the hit. Skinner's intermittent reinforcement remains undefeated for getting organisms hooked to a specific behaviour.

I've had a lot of chats w/folks lamenting the state of peer review at ACL venues. #TIL that @aclrollingreview.bsky.social's Reviewer Guide includes a great intro that we should be passing to new reviewers: aclrollingreview.org/reviewerguid... Not the same as training, obvs, but a good start!

One thing probably clouding CogSci debates about LLMs is that, these debates have deep historical anchors, and when you go back to the times before we all went insane, nobody seems to have expected industry labs to train a model on the entire internet plus custom data as part of a global gold rush.