I agree completely! Some of my work is actually focused on the dataset construction side of this, and I work with materials scientists that do similar (I think?) gaussian process work

It's been really frustrating to see how much this style of "hypothesis generation" work is in AI for science circles. I think there are good ways to use AI in materials science! But so much of the oxygen is taken up by end-to-end, speculative work :/

Between this, the louvre heist, and Charles and Camilla being welcomed to an event with it, what a great year to love dhoom 😍

Based on format alone, sign me up!

...who is this, though 🙈

This is such a neat idea! I wonder how the infrastructure of gating carries over to modalities that are less identity carrying - like, is there an equivalent of this for text? Is there an analogue of signing your input tokens with your private key?

I think it's even more exciting than this!! There's a bunch of stuff conveyed in _not_ text (i.e. diagrams/formulas etc as you point out), but also a lot of stuff contained in _not quite_ text —
formatting, layout, all sorts of fun visual pragmatics that do convey information and are important

👀 very excited to read anything that may come from this 👀

At least some part of this is happening independently! smokesignal.events/about

Fwiw, this is *very* neatly borne out in US military funding solicitations! Control theory really exercised a powerful conceptual hold for a long time

South Asia has plenty of love and use for mayonnaise, just mostly in street food instead of what is considered traditional!

Is Cluely...somehow ethical now?

👀👀👀

I need one of these sized for a cone!!

Womp, wrong hashtag: #acl2025

We argue that models need to be better along four axes: they need to be accessible, personalizable, support iteration, and socially aware.

How might we do that? Come to the poster to find out!

I had so much fun doing this work with Nupoor, @jerelev.bsky.social and @strubell.bsky.social!

Modern tools struggle to support many of these use cases. Even where research exists for certain problems, such as with terminology drift, it requires writing code to effectively use, and is therefore inaccessible, while the accessible tools do not address those use cases at all.

We also saw that they maintained deeply personal methods for reading documents, which led to idiosyncratic, iteratively constructed mental models of the corpora.

Our findings echo early findings in STS, most notably Bruno Latour's account of the social construction of facts!