human user speaks fluent Claude, instance shocked

This would be a good solution. An automatic circuit breaker. If you're getting more than N negative-sentiment replies or QTs, it either sends a notification to you or proactively disables QTs / limits replies. N needs to be adjustable, I think, because it depends on acct size and skin thickness.

A lot of people quote-posting this are condescending to Ian while simultaneously revealing that they don’t understand the publication pace in scientific disciplines. In many fields, you must do triage. You physically could not read everything if you did nothing else 8 hrs/day, 5 days/week.

this is a REALLY good piece and i agree with basically all of it (as a Foucault/Latour person who has published on them both)

Whenever I start to think Bluesky is bad, I remember that brigading on old Twitter got way more personal, with a junior-high-school “let’s ruin her life” vibe that I rarely encounter here. Medievalists basically had to preface every post with “I, for one, welcome our public struggle sessions.”

Ah, nothing says “the holidays” like paperwhites, with their unmistakable fragrance of freshly unwrapped gelatinous lemon air freshener

Solidarity with Jess and Eugene; I've had the experience they're describing. But I continue to be a pollyanna about this. I think it's easy to have substantive AI discourse here -- sharing papers, etc, I never get harassed. But if you do meta-discourse, yes, you've got to set followers-only.

Look, there are a lot of normal people hanging out at the torchlight parade and book burning. You make it sound like we’re all fascists, but we’re just there because our friends are there.

New pictures near the exit of immigration/customs at O’Hare make very f——g clear that you are passing from federal jurisdiction and into the city of Chicago

I think a key problem is that to most critics, it must be everything. It has to destroy the environment, eliminate every job, enable a fascist takeover, melt our brains, all while being a uninspired, unremarkable piece of boring technology that also doesn't work.

I'm on my knees

This program brought together such a wonderful group of ~30 PhD students for two weeks at Berkeley last year with backgrounds in sociology, information, law, computer science and more — same idea but in Oxford next summer; apply by Feb 2 if you’re looking to make connections across fields

Being active on social media either gives you superhuman patience, or makes you go completely insane. No in between.

14 two-year postdocs for academics of any nationality who cannot continue their research due to US politics. Do share if you know of such.

Encoder-decoder models are great for annotation tasks, where you want full information about an input as well as the ability to generate arbitrary structured output. Encoders (BERT) and Decoders (most everything else) only give you one of those.

Just spent an hour wandering through the Kyungpook National University Museum, where labels are mostly Korean-only. I started by asking a model to translate them, but it was absolutely desperate to add commentary to make sure I understood everything in context and why it mattered.

extremely funny how many conversations are people who are being automated out of a job (software engineers) defending ai against people who are not (creatives)

Reminding everyone for no particular reason that Section 230 is one of the last things standing between free speech online and Trump having control over everything you see and say on the internet

THERE'S A MACHINE THAT CAN EXPLAIN NEW THEOREMS IN NATURAL LANGUAGE WHAT IS WRONG WITH YOU PEOPLE. If someone showed you it in 2010 you would literally assume it was faked

Despite those titles, power, and pelf, The wretch, concentred all in self, Living, shall forfeit fair renown, And, doubly dying, shall go down To the vile dust from whence he sprung, Unwept, unhonored , and unsung. ~ Sir Walter Scott

The next edition of the NLP+CSS will be at ACL 2026! It includes an open-ended shared task (work with the Opioid Industry Documents Archive) with travel grants as prizes!

I got wonderful news: I was granted a position of Max Planck independent group leader at the MPI for Empirical Aesthetics ⬇️ (@ae.mpg.de) in Frankfurt, which I'll join in early 2026. It's a huge honor: being trusted with academic freedom offered by MPG. My group will study the evolution of arts. 1/3

I think this is right—and also think it makes it increasingly important for us to figure out how to describe and convey the needed metacognitive skills.

Very happy to introduce a new tool, BookReconciler! You can take spreadsheets with book data and add subject headings, descriptions, ISBNs, HathiTrust IDs, & more. You can also cluster editions & variations of the same "Work." Led by @thisismattmiller.com and supported by @post45data.bsky.social.

People are willing to admit that AI has niche uses (Assyriology &c). So I wonder if Stone Soup would be the smart strategy. “I agree: AI is useless except for a few niche tasks, like coding.” “Glad you came to yr senses.” “And apparently doctors find it useful for paperwork.” “Sure.” “Oh, and …”

it's so refreshing when u find somebody that can take a hint & u don't gotta spell everything out. like u can just say "i like pancakes" and they say "so you hate waffles?" and it's like. YES. THANK YOU. you get me!

Introverts need an ad campaign to explain that we’re not just taking the subway because we’re cheap, but because taxis and ubers involve rapid unscripted human interaction

this is related to what the "Never AI" folks find uncomfortable. We will, I swear, never be able to teach Akkadian to enough students to translate the tablets we have....so what now

How are researchers actually using large language models today, and who benefits most? In our latest WiAIR episode, we speak with Dr Maria Antoniak (@mariaa.bsky.social) about a large-scale survey on LLM use across academia. 👇 (1/6 🧵) #WiAIR

Breaking out of the cold start. We are one year into our 25-year mission to build the Thesaurus of Ancient Greek Numismatics. www.imagines-nummorum.eu/en 🧵 1/3 #RSEng #Numismatics #DigitalHumanities

Like the middle classes, close reading is always rising

🚨NEW publication: "Beyond DeepSeek: China's Diverse Open-Weight AI Ecosystem and Its Policy Implications" How can we understand China's open LLM ecosystem one year after DeepSeek? Our Stanford team finds a rich and varied set of actors and policy implications. hai.stanford.edu/policy/beyon... 1/

New Publication: "Understanding Conversational AI: Philosophy, Ethics, and Social Impact of Large Language Models" (270 pages, Ubiquity Press, open access). Feel free to read it and share it widely! www.ubiquitypress.com/books/m/10.5...

The whole point of buying socks in a big multipack is they should all be identical and interchangeable. But instead, they add a little coloured stripe with just 2 of each colour so you still need to pair them specifically! Stop insisting on sock monogamy you psychopaths!

Surprisingly rapid & high Ai adoption by doctors: 67% use it daily, 84% says it makes them better doctors, 42% says it makes them want to stay in medicine more (10% said less). A lot of the use cases appear to be administrative and research assistance. 2025-physicians-ai-report.offcall.com

2026 abstract submissions are open! ✔️ Submission deadline: March 3rd, 2026 ✔️ Acceptance Notification: April 14th, 2026 View the guidelines here: lnkd.in/eJEXvd8u

Daily affirmation

Yesterday I told someone “I’m going to steal that but credit you,” so this is me crediting: Tai-Won Kim at Sogang University has students write a short essay by hand on an in-class exam, but then also later expand it into a polished draft using both AI and human feedback. I like that combination. +

Use AI unto others as you would have them use AI unto you

I presented our poetry research at this seminar, featuring work in collaboration with @bhyravajjula.bsky.social, @mellymeldubs.bsky.social, and Anna Preus. What can we learn about poems using LMs? What can we learn about LMs using poems? PDF: drive.google.com/file/d/1Cu0H...

Does anyone find that your brain enters babytalk mode after a day of immersion in an unfamiliar language? I start obsessively muttering the 2 or 3 characteristic phrases I’ve learned, like my brain thinks I can become German if I just say “doch” enough.

LMs need linguistics! New paper, with @futrell.bsky.social, on LMs and linguistics that conveys our excitement about what the present moment means for linguistics and what linguistics can do for LMs. Paper: arxiv.org/abs/2501.17047. 🧵below.

I always learn so much from @mariaa.bsky.social. She's really helped me think more deeply about AI models and privacy, which sometimes doesn't get enough attention. It's like social media all over again, but now the data is even more sensitive. This clip is worth a listen!

Individualist framing — “Are your microblogs changing voters’ minds?” — is kind of a red herring. Sure, the answer to that question is “no.” But Mike’s point here is more important: communities do matter, and it matters that we create the ones we want to see.

The audio recording of our conversation from my book launch with Ted Chiang and James Yu is now available! I present Artificial Humanities, James runs a demo of using models for creative writing, and then we discuss burning questions around AI and writing: www.commonwealthclub.org/events/archi...

If the timing of my activity here seems off, it’s because I’m in another time zone for the week, learning about digital humanities research in Korea.

New article in AI & Society with @richardjeanso.bsky.social and @hoytlong.bsky.social 🎉 We wanted to know how AI might affect cultural fields like literary publishing. But cultural production is complex! So we piloted a new method we call “social simulation.” rdcu.be/eTkMy

to my great surprise, we are left with some money for buying books for our team at the institute. Any recent recommendations in computational humanities, DH, poetics, verse studies, world lit / comparative lit? Let's go, internet! Self plugs are more than welcome.

Introducing Bolmo, a new family of byte-level language models built by "byteifying" our open Olmo 3—and to our knowledge, the first fully open byte-level LM to match or surpass SOTA subword models across a wide range of tasks. 🧵

Texts from my husband who is shopping. Maximus: Smells like performed masculinity!