Lightnews — Scholar-powered news

Reposted by Jacob Eisenstein is at CoLM

Maria Antoniak @mariaa.bsky.social · 2d

Here’s a #COLM2025 feed!

Pin it 📌 to follow along with the conference this week!

Reposted by Jacob Eisenstein is at CoLM

Myra Cheng @myra.bsky.social · 5d

AI always calling your ideas “fantastic” can feel inauthentic, but what are sycophancy’s deeper harms? We find that in the common use case of seeking AI advice on interpersonal situations—specifically conflicts—sycophancy makes people feel more right & less willing to apologize.

Screenshot of paper title: Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence

2 38 96

Reposted by Jacob Eisenstein is at CoLM

Pete Shaw @ptshaw.bsky.social · 7d

Excited to share a new paper that aims to narrow the conceptual gap between the idealized notion of Kolmogorov complexity and practical complexity measures for neural networks.

Bridging Kolmogorov Complexity and Deep Learning: Asymptotically Optimal Description Length Objectives for Transformers

1 5 8

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Aug 27

🛸

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Aug 26

thanks! i was more confused about the “kugel” part but TIL that this is apparently inspired by an airy globe?

1 1

Reposted by Jacob Eisenstein is at CoLM

Joshua קאַבאַק-געװירץ Raclaw @joshuaraclaw.com · Aug 25

Cannot stress enough how good it is that you can come across a post about gorgeous little Yiddish book sitting in someone’s family collection, and within a few seconds you can find the full scanned version of the book available for free through the Yiddish Book Center’s website

4 7 44

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Aug 25

1 6

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Aug 25

i think my great grandmother was the last owner of these books that knew how to read them

1 5

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Aug 25

found some books at my parents’ house

automatic translation: autonomy by dr. b hoffman

2 1 20

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Aug 19

1 6

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Aug 11

On the positive side, this vario grinder, which i bought second hand, is the best technological upgrade of the summer in my house.

(Its grind settings are 1-10, a-z, so the chatgpt output is clearly wrong and the claude output is nonsensical)

2

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Aug 11

Baristas still safe from robotic automation, and not just because robots don’t know what coffee tastes like.

prompt: “I’m trying to dial in this v60 of huatusco with my vario. temp / grind recommendations?”

1 3

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Aug 8

right but you’ll notice it’s pretty hard to validate a proposed answer to those why questions, so it was not unreasonable to hypothesize that a better formal model of language might yield better features for an NLP system

2

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Aug 8

The project of putting statistical meat on grammarian bones is, imo, a beautiful one (this is basically what textbook is about), even if it didn’t work out as a way to build NLP. It was helpful to have the participation of people like Bender, who understood the latest ideas in theoretical syntax.

9

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Aug 8

Bender is/was fairly distinct among syntax people for caring about statistical NLP and for believing that it can or even must incorporate sophisticated ideas about grammar. Lots of NLP people thought this for a long time, but I don’t think many linguists did.

3 6

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Aug 8

I’d guess that the majority position of syntacticians about LLMs (and other NLP beforehand) is roughly what Chomsky says: language tech can’t possibly teach us anything about the human language capability, so whether the LLM writes well doesn’t matter at all.

rev. howard arson @theophite.bsky.social · Aug 8

the thing about Emily Bender is that the reason she hates LLMs is that the fact that they exist -- not that they are "AGI," but the fact that they exist -- falsifies every paper she has ever written. you more or less can't be a formal grammar person in a world where statistical learning works.

1 10

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Aug 7

boston champaign pittsburgh atlanta, and, uh, let’s count seattle

glad i did it, hope i don’t have to do it again

GHOULLIKEHELLMACHINE @golikehellmachine.com · Aug 7

under golikehellism everyone will be required to live in at least three new cities where they know less than five people from the age of 20-40

1 2

Reposted by Jacob Eisenstein is at CoLM

Margaret Mitchell @mmitchell.bsky.social · Aug 6

🤖 ICYMI: Yesterday, @hf.co and OpenAI partnered to bring open source GPT to the public. This is a Big Deal in "AI world". Allow me to explain why. 🧵
huggingface.co/openai/gpt-o...

2 17 54

Reposted by Jacob Eisenstein is at CoLM

Ibn Bassal @ibnbassal.bsky.social · Jul 29

roman burrito thread

Ed @ed3d.net · Jul 26

you also have the laganum, which is (at some points) a pancake-ish kinda dough snack. slather on a "salsa" of garlic, onions, olives, and vinegar. pile on some roasted lamb or pork and caramelized onions, with some parsley or cilantro. chickpeas sub for beans. artichoke hearts, too

3 36 140

Reposted by Jacob Eisenstein is at CoLM

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Jul 29

this is very cool and i’m looking forward to reading the paper, but a basic question about this data: isn’t it likely that a congressional rep’s speeches are written by a shifting cast of speechwriters over the course of their career? wouldn’t that explain adoption of new usages?

1 1 3

Reposted by Jacob Eisenstein is at CoLM

Gaurav Kamath @grvkamath.bsky.social · Jul 29

Our new paper in #PNAS (bit.ly/4fcWfma) presents a surprising finding—when words change meaning, older speakers rapidly adopt the new usage; inter-generational differences are often minor.

w/ Michelle Yang, ‪@sivareddyg.bsky.social‬ , @msonderegger.bsky.social‬ and @dallascard.bsky.social‬👇(1/12)

3 17 33

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Jul 29

this is very cool and i’m looking forward to reading the paper, but a basic question about this data: isn’t it likely that a congressional rep’s speeches are written by a shifting cast of speechwriters over the course of their career? wouldn’t that explain adoption of new usages?

1 1 3

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Jul 25

which, if it wasn’t obvious in 2024, should be hard to miss in 2025 www.whitehouse.gov/presidential...

Preventing Woke AI in the Federal Government

By the authority vested in me as President by the Constitution and the laws of the United States of America, it is hereby ordered: Section 1. Purpose.

www.whitehouse.gov

1 2

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Jul 25

Readers interested in bias and local knowledge may be disappointed by the thin discussion in section 5: the PPI work is great, but if Jordan means "bias" in anything like the colloquial sense then it's impossible to talk about without reference to power differentials
arxiv.org/abs/2005.14050

Language (Technology) is Power: A Critical Survey of "Bias" in NLP

We survey 146 papers analyzing "bias" in NLP systems, finding that their motivations are often vague, inconsistent, and lacking in normative reasoning, despite the fact that analyzing "bias" is an inh...

arxiv.org

1 5

Jacob Eisenstein is at CoLM @jacobeisenstein.bsky.social · Jul 25

Indeed, while social situations can *create* new kinds of uncertainty -- e.g., uncertainty about others' intentions -- they can also create powerful incentives for agents to reason and communicate about uncertainty. (Shameless self-promotion here: arxiv.org/abs/2503.14481)

Don't lie to your friends: Learning what you know from collaborative self-play

To be helpful assistants, AI agents must be aware of their own capabilities and limitations. This includes knowing when to answer from parametric knowledge versus using tools, when to trust tool outpu...

arxiv.org

1 3