David Mimno

I love "low stakes programming"! The characteristics seem to be: highly task focused, narrow user base, easily replaced. This sort of thing is hard because it's often interaction-oriented, requiring lots of complex library/framework interactions.

Microsoft Research NYC is hiring a researcher in the space of AI and society!

Screenshot of the Open Visualization Academy's website, showcasing our logo and all courses available tomorrow

The Open Visualization Academy launches TOMORROW MORNING (EST) openvisualizationacademy.org with 7 courses; we'll release 1-2 more/month beginning in February.

Sneak peek at our website, designed by @vsueiro.bsky.social and Melissa Strong #dataViz #dataVisualization #infographics #dataJournalism

I made a snarky Western list to show how far behind we are in fostering an ecosystem vis a vis China.

‘Red team’ students stress-test NYC health department’s AI | Cornell Chronicle People usually strive to be their true, authentic selves, but this fall, five master’s students at Cornell Tech adopted not only alter egos but also “bad intent,” in an effort to make AI safer for hea...

Hello!

This is a reminder that @cornelltech.bsky.social runs a Red Team Clinic that provides a *free* safety consultation to nonprofits / public sector orgs that are developing a public-facing AI tool and want to stress-test it for possible abuse vectors.

Applications welcome on a rolling basis:

Wow, this is spectacular! Evolution of beak size in an urban population of juncos in response to the loss of supplementary food (bird feeders) during the COVID pandemic, and subsequent regression. Is the UCLA campus the new Daphne Major?
pnas.org/doi/full/10....

A colleague pointed out that if you ask an LLM to rate docs one by one, it will give them all middling scores. But if you ask it to rate them all at once, it differentiates. Given that a lot of reviewing is quietly AI, could we explicitly design a stronger, better-calibrated AI review phase?

For those not into book history, that means it is impossible that even a single person had ever read the rest of it.

The remains are probably still on a shelf at Perseus HQ, with rubber bands holding them together. I think about it sometimes. I'm not sorry.

This English translation of Pliny the Elder's Natural History exists online because I sliced the spines off a beautiful four-volume edition from 1850-something in ~2003. In the previous 150 years only the quires for the first 100 pages had been cut.

www.perseus.tufts.edu/hopper/text?...

📒 Wrote a quick blog post and Colab notebook showing how to use ai.dev 's batch mode with @hf.co Datasets, Google Cloud, and one of our Gemini models (Gemini 2.5 Flash Lite, but you could swap in any model docstring!).

🔗 dev.to/googleai/ben...

🔗 github.com/dynamicwebpa...

I wanted to note that I made a list of all ancient comets from Med antiquity (500 BCE-400 CE) for class a number of years ago. Here is my excel sheet, based on Ramsay, Descriptive Catalogue of Greco-Roman Comets from 500 B.C. to A.D. 400. Syllecta Classica, 17. 2006.
docs.google.com/spreadsheets...

Come work with us! Tallinn Uni is hiring a second lecturer in DH&AI. Estonian proficiency not required, foreign applicants welcome, required phd can be in progress (expected to be completed), competitive salary+benefits, nice working environment close to the sea
candidate.recrur.com/public/jobad...

Ah, that makes more sense!

en.wikipedia.org/wiki/Siege_o...

What’s the temporal limit? Nothing in Greece south of Thessaloniki is weird

The seven layers are:
Moral Source: where authority comes from.
Epistemic Posture: how certainty and error are treated.
Verification Logic: what counts as evidence.
Deliberative Logic: how disagreement is bounded.
Accountability Direction: where consequences flow.
(cont)

"Abolish ICE" is the law-and-order position

Requires numba to be fast, and doesn't yet implement sampling tricks, but already competitive with VB and higher quality

GitHub - mimno/sklearn-gibbs-lda: sklearn-compatible LDA using collapsed Gibbs sampling with Numba optimization sklearn-compatible LDA using collapsed Gibbs sampling with Numba optimization - mimno/sklearn-gibbs-lda

Drop-in replacement for sklearn LDA using MCMC and hyperparameter optimization: github.com/mimno/sklear...

Thanks!

Entire Eastern US planning to experiment with powdered water this weekend ❄️

Why are citations so hard? LLMs are interpolation machines. It works well for, say, programming questions: rather than parsing 3-4 not-quite-relevant Stack Overflow posts, you can get an answer to your exact question. But doing the same trick with refs doesn't work, they either exist or don't.

Are you a student interested in a transformative summer experience? Cornell Tech’s Summer Innovation Intensives (July 13–30) blend college-level coursework, real-world projects, and NYC immersion.

Info Sessions:
Jan 25 (in person)
Jan 28 (virtual)

RSVP: https://bit.ly/3NhhNnQ

Are you willing to accept the consequences if those rocks are ever released into the wild, Scott? These are the worst of the worst.

Professor Kate Starbird named ACM Fellow for advancing research on misinformation and information ecosystems HCDE Professor Kate Starbird has been named an ACM Fellow, one of the highest honors in computing, for her groundbreaking research on misinformation and online information ecosystems. The honor recogn...

HCDE Professor Kate Starbird has been named an ACM Fellow, one of the highest honors in computing. She's recognized for contributions to understanding and improving information ecosystems, including during crisis events and addressing misinformation. www.hcde.washington.edu/news/article...

PhD student in Computing Science with a focus on Machine Learning

We are #hiring a #phd student to develop novel #Bayesian #MachineLearning methods and algorithms. A fully funded PhD position at @umeauniversitet.bsky.social. See full ad and apply here: www.umu.se/en/work-with...

Citation is the foundation of academic promotion. It’s noisy, sure, but its integrity is worth fighting for. Hallucinated citations should be a desk reject.

GitHub - mimno/translategemma: Latin to English translator using TranslateGemma via Ollama Latin to English translator using TranslateGemma via Ollama - mimno/translategemma

Claude Coded a command line translation wrapper for ollama translategemma, with LA->EN as default. It works really well!

github.com/mimno/transl...

TranslateGemma: A new suite of open translation models TranslateGemma is a new family of open translation models built on Gemma 3.

Meet TranslateGemma. 💎
​✅ Open weights (4B, 12B, 27B)
✅ 55 languages + 100s more in training data
✅ Multimodal capabilities (image text)
Blog: blog.google/innovation-a...
Paper: arxiv.org/pdf/2601.09012
Model: huggingface.co/collections/...
Cookbook: colab.research.google.com/github/googl...

I'm trying to find scholars who have studied how scientists assimilate the discoveries of other scientists. I'm trying to understand the potential impact of AI-assisted acceleration of scientific discoveries beyond the short term. Can anyone point me to people and publications?