Profile avatar
juand-r.bsky.social
CS PhD student at UT Austin in #NLP Interested in language, reasoning, semantics and cognitive science. One day we'll have more efficient, interpretable and robust models! Other interests: math, philosophy, cinema https://www.juandiego-rodriguez.com/
376 posts 4,030 followers 2,502 following
Prolific Poster
Conversation Starter

“Pretty much wherever you are, you’re probably less than 30 minutes away from your closest No Kings event.” - @leahgreenberg.bsky.social with @democracynow.org We hope to see you at your local #NoKings Day mobilization on Saturday: www.nokings.org?SQF_SOURCE=i...

Looking forward to the preprint!

I'm training an SVM to classify correct/incorrect proofs of suitably encoded mathematical proofs, such as Fermat's Last Theorem's. Unsuccessfully so far, the margin is too small

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 think what’s most dangerous about this belief is that it legitimizes the structural demolition of science the US is carrying out and its underfunding elsewhere

This in-depth documentary from Samuel Black at @moreperfectunion.bsky.social on xAI's Colossus supercomputer in Memphis, TN for training Grok is a phenomenal example of the kind of in-depth AI accountability reporting we cannot have enough of. www.youtube.com/watch?v=3VJT...

"Large [language] models should not be viewed primarily as intelligent agents but as a new kind of cultural and social technology, allowing humans to take advantage of information other humans have accumulated." henryfarrell.net/wp-content/u...

One striking thing from this (excellent) report on a DOGE employee's use of AI is that everything about it is antithetical to sound software engineering practices—an engineer hurriedly deploying an unfamiliar tool to meet an impossible deadline for a problem domain in which he has no expertise.

I’m excited to start my summer research internship at Abridge!

togelius.blogspot.com/2025/06/the-...

News🗞️ I will return to UT Austin as an Assistant Professor of Linguistics this fall, and join its vibrant community of Computational Linguists, NLPers, and Cognitive Scientists!🤘 Excited to develop ideas about linguistic and conceptual generalization (recruitment details soon!)

I can’t believe I only just discovered Beer.

psycnet.apa.org/record/1994-...

Hoy, en esta isla, ha ocurrido un milagro

Those who can speak out without fear of retribution (e.g., losing their visa, green card, or getting thrown in a van by goons) should protest

Revoking visas to Chinese students in the US is both cruel and stupid. Immigrants and investment in science made this country great. They are throwing it all away for no reason

"Tell, Don't Show" was accepted to #ACL2025 Findings! 

 Our conceptually intuitive, lightweight approach for literary topic modeling combines the new (language models) with the old (classic LDA) to yield better topics. ✨📚 arxiv.org/abs/2505.23166

The more leches the better

"The AI industry is saturated in spectacle. The spectacle of AI, more so than the tool of AI, speaks to dreams of generalities, born from abstractions, forced together through the violent reduction of our world to tiny signals." - @eryk.bsky.social mail.cyberneticforests.com/human-noise-...

It's an exciting prospect to have corporations provide AI friendship and companionship services, competing with legacy friendships and seeking to establish a friendship monopoly. Software eating yet another aspect of the world

Check the bonus panel on the site! ◆ COMIC ◆ www.smbc-comics.com ◆ PATREON ◆ www.patreon.com/ZachWeinersm... ◆ STORE ◆ smbc-store.myshopify.com #webcomic ⁠smbc

Too many interesting papers coming out, too little time

It's good to finally have a good reference for this stuff! Kudos to the authors. arxiv.org/abs/2501.18374

www.thecrimson.com/article/2025...

Most of us in higher ed in the US are going to lose something over the next four years (opportunities, or funding, or time). But if we believe universities exist, in part, to preserve historical memory and create a space for dissent, our primary job right now is to fight and take the hit.

The fascist spectacle is the point.

Most researchers don’t believe AGI is coming any time soon. But policy makers are steering policy toward AGI anyway. In this article for Tech Policy Press, a look at the distortions the AGI Frame introduces to policymaking. #ai #criticalai #aipolicy In @techpolicypress.bsky.social

WTF

defector.com/close-readin...

This is very cool arxiv.org/abs/2505.13755 x.com/wgilpin0/sta...

US NSF funding cut in half Details of cuts to US National Science Foundation (NSF) www.nytimes.com/interactive/...

choose your adventure

Introducing SimpleStories: A synthetic story dataset and model suite designed for understanding the internals and learning dynamics of LMs. It's an evolution from TinyStories and leverages better LMs for data generation and offers more data diversity. 🧵

You may know about Bell primes, Eisenstein primes or Fermat primes, but did you know about the Dickens primes? They are the best of primes. They are the worst of primes.

Why does this technology exist?

An actually useful dialogue about things like "should AI take care of my personal chores/comms/etc" should divorce the subject from AI as a technology and think about mechanization more generally mitpress.mit.edu/978026252652...