You have a moral imperative to refuse to work with these people or develop models for these purposes.

*CL folks, a recent history question: was the move to OpenReview as the review platform for *CL conferences related to the move to ARR? Or did they just happen simultaneously? Asking for a discussion with @shaily99.bsky.social

I'm really not a fan of the way some journals in the physical sciences state results upfront, and bury methods much later in the paper, or even in supplementary materials. Maybe it's my ML bias, but I don't trust that your methods are prima facie reasonable!! Show me what you did before the results!

Appointment reading

I think my AI beliefs are quickly becoming ‘this has enormous positive potential if implemented responsibly. it will not be implemented responsibly and like maybe five companies are even trying.’

if quantitative social science is automatable, it's because many of the fields in question have already reorganized their research in such a way to be indistinguishable from machine output, and already optimized for machine measurement.

This is a great thread, and makes me think about how the affordances of LLMs (largely text-only chat) preclude the kind of epistemic vigilance we were able to develop around tech like search. I think we can build LLM-based tech that's good here, but chat ain't it

Rename LASSO and Ridge to kiki and bouba

This does feel like situation where open models are held back by the state of benchmarks. For an open model, there isn't as much of a theory of use outside of benchmarks. For closed models, the companies putting them out can build for business cases and their associated internal benchmarks

Most tech workers, even at Palantir (probably), are pro-social compared to the tech owner and VC classes. They need unions so their voices matter for company governance. www.wired.com/story/palant...

Vivid mental image of cellino and barnes being like penn and teller, where one is always completely silent and if you advance a surprising legal theory your consultation is free

Firewatch is one of the most lovely games I've played. I've taken so much joy from playing it with multiple friends and seeing how different a game it is for them. I didn't know games could tell stories like this before it (It also played a small part in my fiancée and me getting together!)

This is a real banger of a paper. The example of a model being weirdly focused on jasmine (lol) makes me increasingly think that single-point-of-access models don't really consider who their audience is. Jasmine is a super legible cultural marker for people outside, but is so, _so_ generic.

“Position paper” is just a label to make some kinds of interdisciplinary theoretical work fit into the CS publishing schema.

extremely on brand, love this town

Best use of this meme i’ve ever seen

Another trip to Bangalore, another time that Google Maps (but no other Google service) turns itself Indian accented and maybe can't be turned back? 🤔

This is very millennial of me, but I really want a buzzfeed-style article summarizing the gas town discourse

This is a really interesting paper, and there's an interesting structural angle to the narrowing of focus. ML approaches involve building an infrastructure - data, models, etc. Once that infra exists, it's easy to apply it to similar problems - leading to the kind of clustering described here...

"We should keep build-and-test empiricism—its speed, openness, and compounding progress are real achievements—but we should supplement it with a second language of empiricism that lets us state and evaluate claims with more precision."

Researchers: Aside from moderation and safety features, what would you want in a researcher-focused social app? Obviously a feed of shared papers, but what else?

The two hardest problems in Computer Science are 1. Human communication 2. Getting people in tech to believe that human communication is important

This is an excellent piece on the way that a culture of repurposed and thin measurement has really warped what it means to do science. The risk of AI here is not just intensifying the use of these metrics, but reifying those flawed standards into systems that make the culture harder to change.

As a PhD student in computer science, the kind of knowledge production that is being "innovated" away here is exactly what departments like mine could use a lot more of. This is a real loss for CMU.

"the constraint is the feature" is an excellent way to put this. General purpose tools are only general purpose until you use them for something!

What if technology didn’t feel so… hollow? Some friends and I just released a manifesto about a world where tech leaves us feeling nourished (along with an evolving list of theses about how we can build it) resonantcomputing.org

A NeurIPS feed! Pin it and follow along.

It's strange to be in a field wishing for the bubble around you to pop, but I think this post captures some of the why. AI as technology has so much cool and fun potential, and it's going to be so exciting to see it freed from being developed and pushed by the awful people at the top of tech.

This is so exciting!! As someone that works on scientific PDFs, the extent to which format constraints limit how we represent the complexity of the science we do is maddening. I hope other venues adopt things like this soon!

i don't agree with the attitude of some faculty that their job is to train people who can run the code, even if they don't understand the underlying physics i want a society of people who are trained to be curious about the code they're being asked to run and capable of understanding it, actually

Please respond to this survey if you have changed or have thought about changing your name in academic publishing! For any reason, whether it be transition, recognizability, marriage, privacy, immigration, cultural reasons, etc. Please RT for reach :)

love the construction of this dataset, it feels very much of a piece with other work like the reversal curse. LLMs clearly don't "learn" or "know" in the ways we do, whether relation reversal or simple compositionality, and I love how clearly this work demonstrates that.

My name is Ozymandias, King of Kings! Look upon my works, which achieve state-of-the art performance across a diverse suite of benchmarks

✨I’m on the academic job market ✨ I’m a PhD candidate at @hcii.cmu.edu studying tech, labor, and resistance 👩🏻‍💻💪🏽💥 I research how workers and communities contest harmful sociotechnical systems and shape alternative futures through everyday resistance and collective action More info: cella.io

How to not do computational humanities: (1) Lay out a question/hypothesis about a complex, cultural domain. (2) Compute numbers that were inspired by (1) but without sufficiently formalizing (1) so as to meaningfully link to it. (3) Interpret numbers to mean whatever your prior vibes were.

So excited to be TAing this course! So much of the knowledge you have as a PhD student is expected to be gained by osmosis, and it leads to some odd holes and gaps. This course should fix most of that problem!

The military keynesianism of our day (though it's of course possible that military keynesianism is the military keynesianism of our day, and this is just _another_ one)

Coming soon (6pm!) to the #ACL poster session: how do experts work with collections of documents, and do LLMs do those things? tl;dr: only sometimes! While we have good tools for things like information extraction, the way that experts read documents goes deeper - come to our poster to learn more!

Gently, I would like to say: When people tell you that they would appreciate a feature that does something automatically, it's not responsive to that concern to explain that by going through several steps for every individual instance, they can get the same result in each instance.

OpenAI has effectively conned people into thinking that Chatbots & AI "Assistants" are The FEWTCHA of AI. Friends, they are most likely *not.* Neither are the big cloud-based Generative AI services. Small, purpose-fit, on-device models that make your existing activities easier/better? There you go.

When it comes to text prediction, where does one LM outperform another? If you've ever worked on LM evals, you know this question is a lot more complex than it seems. In our new #acl2025 paper, we developed a method to find fine-grained differences between LMs: 🧵1/9

I for one am grateful for the opportunity to meditate on the meaning of “scientific artifact” at 2:15am

Great points in the replies here re:civility being an unequally applied, generally awful standard. It's also a great example of how common formulations of "sentiment analysis" or "toxicity detection" in NLP that are non-contextual lead to systems that stop sounding good with even slight scrutiny.

AI as another pivot to video is a thought I've also had, and this article is such a great articulation. The way "AI" is framed, pitched, and deployed by big players reflects at best ignorance, and at worst a real contempt for the social nature and humanity of our jobs and online spaces

This talk was such a joy to do! If you'd like to read the paper, it's here: arxiv.org/abs/2411.17840. Thank you for having us, @patrickbriansmith.bsky.social!

If you're at NAACL this week (or just want to keep track), I have a feed for you: bsky.app/profile/did:... Currently pulling everyone that mentions NAACL, posts a link from the ACL Anthology, or has NAACL in their username. Happy conferencing!

Ever trusted a metric that works great on average, only for it to fail in your specific use case? In our #NAACL2025 paper (w/ @841io.bsky.social), we show why global evaluations are not enough and why context matters more than you think. 📄 aclanthology.org/2025.finding... #NLP #Evaluation (🧵1/9)

🚀 Excited to share a new interp+agents paper: 🐭🐱 MICE for CATs: Model-Internal Confidence Estimation for Calibrating Agents with Tools appearing at #NAACL2025 This was work done @msftresearch.bsky.social last summer with Jason Eisner, Justin Svegliato, Ben Van Durme, Yu Su, and Sam Thomson 1/🧵

This is absurdly great, but I haven't read a single news article about it. A fully open source, offline-first alternative to Notion that's a collab between the French and German governments because they want to host docs securely and on their own terms. THIS is what Europe should be doing.

one thing in AI is not new -- people taking one small part of a job, mischaracterizing it, ignoring all the other stuff, and then assume the AI can do the whole job on its own