Suresh Venkatasubramanian
geomblog.bsky.social
Suresh Venkatasubramanian
@geomblog.bsky.social

Director, Center for Tech Responsibility@Brown. FAccT OG. AI Bill of Rights coauthor. Former tech advisor to President Biden @WHOSTP. He/him/his. Posts my own.

Suresh Venkatasubramanian is an Indian computer scientist and professor at Brown University. In 2021, Prof. Venkatasubramanian was appointed to the White House Office of Science and Technology Policy, advising on matters relating to fairness and bias in tech systems. He was formerly a professor at the University of Utah. He is known for his contributions in computational geometry and differential privacy, and his work has been covered by news outlets such as Science Friday, NBC News, and Gizmodo. He also runs the Geomblog, which has received coverage from the New York Times, Hacker News, KDnuggets and other media outlets. He has served as associate editor of the International Journal of Computational Geometry and Applications and as the academic editor of PeerJ Computer Science, and on program committees for the IEEE International Conference on Data Mining, the SIAM Conference on Data Mining, NIPS, SIGKDD, SODA, and STACS. .. more

Computer science 89%
Engineering 5%
The Rolling Pebbles
Stationary stone?

Led Balloon
Slightly diminish a band:

Jefferson Kite
Slightly diminish a band:

They Might Be Unusually Large
Suspect arrested in predawn fire that left parts of Mississippi’s largest synagogue in charred ruins

A fire heavily damaged Jackson’s only synagogue before dawn Saturday – the same house of worship that was firebombed by the Ku Klux Klan in 1967 because the rabbi had been an advocate for civil…
Suspect arrested in predawn fire that left parts of Mississippi’s largest synagogue in charred ruins
A fire heavily damaged Jackson’s only synagogue before dawn Saturday – the same house of worship that was firebombed by the Ku Klux Klan in 1967 because the rabbi had been an advocate for civil rights.
mississippitoday.org
Slightly diminish a band:

Jefferson Kite
Slightly diminish a band:

They Might Be Unusually Large
Slightly diminish a band

Something Something
ACLU @aclu.org · 2d
We released a new report in partnership with the Center for Tech Responsibility at Brown University on how policymakers and researchers can better analyze AI legislation to protect our civil rights and liberties.
Making Sense of AI Policy Using Computational Tools | TechPolicy.Press
A new report examines how to use computational tools to evaluate policy, with AI policy as a case study.
www.techpolicy.press

How much is Slack paying you :):)

is it a fancy LLM? and is it something that provides useful and relevant feedback? how is it being trained to do that? and is there a non-deli paper that describes said process of training?

I don't know if it is. Presumably the ICML organizers do. The structure of these interventions is basically to address "authors as DDOS attack on ICML" :), with the hope that this will free up reviewers.

The new policy on deli papers (aka thinly sliced:)) seems quite strong and new. I wonder how it will play out. In brief, authors have to disclose all papers they are submitting as prior work to the one under review. I'm also not encouraged by the "use our LLM to review your work"

I am seen.
Here are some otters hammock hugging - Enjoy!
some citation graphs data pipelines will create new "paper" nodes based on extracted bibstrings from PDFs

so in 2026 the papers we hallucinated in 2025 might end up being "real" papers on gscholar or sthn lol

given the way fast bowlers are breaking down, I'd guess they are NOT using PEDs because that's one value add there - for recovery from injury (unlike the EPO stuff that seems cycling specific)

This is beginning to sound like an episode of Wait Wait Don't tell me where you have to identify the true news story from the fake ones :)

Not sure why that would help. Some amount of adrenaline is a good thing for a batsman :)

One aspect of SVMs that are enduring is the idea of high dimensional representations that allow for simple classifiers. You see a lot of that emerging in LLM land as well.

Yes. A standard measure might be the ranking distance (number of swaps needed to go from one order to the other). I don't have data on prior years though.

Why it's hard to predict NFL standings. This is a comparison of preseason and current standings.
Oh, but did you know that b/c of the number-theoretics of 2026, the year allows us to give a dynamical reformulation of the P vs NP problem (A Millennium Prize Problem, worth $1M), thereby showing both that it is the wrong question to ask and that P is different from NP. (10 pages, written with AGI)

Oof. At least they should get the definitions right.
I'd like to propose the following norm for peer review of papers. If a paper shows clear signs of LLM-generated errors that were not detected by the author, the paper should be immediately rejected. My reasoning: 1/ #ResearchIntegrity
NYC Mayor-elect Zohran Mamdani on Mustapha Kharbouch, the Palestinian student who was falsely accused last week of being the Brown University shooter by various racists and conspiracy theorists.
In the late 1830s, Karl Weierstrass dropped out of university. He is said to have spent his school years drinking and fencing. Decades later, he published a function that threatened everything mathematicians thought they understood about calculus. www.quantamagazine.org/the-jagged-m...
Really enjoyed this roundup of fun animal stories from 2025 🧪 And the bone-collector caterpillar is still my favorite!
These are our favorite animal stories of 2025
From clever cockatoos to vomiting spiders, these cool critters captivated us this year.
www.sciencenews.org
The full spiked 60 Minutes CECOT package, clean & subtitled. 1/5
Very accurate observations. ML and petter recognition were not regarded as mainstream AI from the late 1960s through early 1980. The ML community had trouble getting thei papers accepted in AI journals.
Wildly different things, tasks, techniques, subspecialties being lumped into "AI" and then being conflated with each other, doesn't help. Different types of models vs the techniques to train them vs the tasks they are supposed to accomplish, all under "AI".

this is the way. Next thing he'll be calling you 'Clem-o' :)