.
@rajivmovva
,
@PangWeiKoh
, and I write for
@NatureMedicine
on using unlabeled data to improve generalization + fairness of medical AI models:
We highlight two nice recent papers illustrating this - , .
I am recruiting PhD students for Fall '24! Please apply to Cornell CS if you are interested in ML, data science, health, or inequality. Feel free to retweet!
We are based in NYC - here's the view from our island campus (taken during Pride - note the rainbow lights!)
man, I started watching Tim Roughgarden's lectures because I wanted to learn a bit of algorithmic game theory ()
but I may keep watching just to learn how to teach. I'm not sure I've ever seen a stronger opening to a class.
I am very excited to announce that I will be starting in summer 2021 as an assistant professor in computer science at Cornell Tech, Cornell's new NYC campus () and that I'll be spending the next year as a senior researcher at Microsoft Research New England!
Hello! I'm recruiting students interested in ML/data science, especially as applied to healthcare or inequality. If this is you, please consider applying to 1) MSR internship ( - for current PhD students) and 2) Cornell CS PhD. Please RT.
(please retweet) I am taking CS PhD students! If you are interested in ML, data science, health, or inequality, please apply to Cornell :)
We have an island campus in New York City and a fantastic group of students I feel lucky to collaborate with every day. Join us!
I won an NSF CAREER award to study "Equitable medical decision-making"! I learned it was recommended for funding the day before my birthday :) Huge thanks to my collaborators and Cornell mentors who taught me how to write my first NSF grant. Beyond work, I wanted to say... 1/2
I may start a Twitter account entitled "Computer Scientist Confused by Conventions in Econ Papers". For example: what's with writing "[Figure 1 about here]"? Just put Figure 1 there! It will be fine!
Honored to have been named to MIT
@techreview
's 35 Innovators Under 35. Very grateful to the mentors and friends who collaborated on the work this award recognizes!!!
Faculty who mistreat PhD students do so in part because they think they can get away with it
But as one small datapoint - if I learn someone treats students badly, it tanks my willingness to work with them. Don't care how famous they are. It infuriates me to see this stuff. 1/2
I have a postdoc position available in machine learning for healthcare, health equity, and public health! Come work with us at the beautiful
@cornell_tech
campus in New York City. Please reshare :)
My PhD class, "Data Science for Social Change" - - discusses 1) how to do impactful data science and 2) how to write about it for a mass audience.
I am very excited! We get to read many of my favorite papers and talk to a ~dozen of my favorite writers.
I've never found a calendar app I like, so I wrote a lil Python script today which logs into all my Google calendars, computes my free blocks, and prints them out for scheduling emails...let me know if this already exists, otherwise maybe I'll post it on Git if it's useful.
Many people are discussing how Turing Award winner Jeff Ullman used to have a webpage making offensive comments about Iranians and Native Americans, but no longer does. I want to give a shoutout to a person who played a small role in getting the page taken down,
@PangWeiKoh
. 1/
My boyfriend’s parents started reading this book on algorithms and then my parents started reading it too and now my mom is talking to my boyfriend about merge sort and thank goodness my vacation is ending soon so I can get a break from all this CS.
I was very happy to receive
@cornell_tech
's Faculty Teaching Award, given to the faculty member students believe has most advanced their educational experience. The best surprise in my first year has been how much I like teaching, and going beyond the math to tell human stories.
hello
@MSFTResearch
I am pleased to report that on a Sunday 8 months after my start date I have paid my first visit to the office
could someone come out and onboard me please it is quite cold
Hi research folks :) I am wondering when you tend to get ideas for projects. Do you tend to get ideas while going for walks? While reading papers? While talking to peers? While playing with your dog? While trying to think of new ideas? Trends/anecdotes welcome, feel free to RT.
Bad news, guys - I have moved to a new apartment and am morally obligated to bombard you with yet more sunset pictures from a slightly different angle. Very excited to watch progression of sunset angle over spring and summer months!
#RooseveltIsland
Sister defending her PhD dissertation today, then girlfriend giving a job talk - the whole morning is pretty much a movie montage of academic women doing great things.
Excited to announce the our new Coursera course, "Practical Steps for Building Fair Algorithms", taught with
@__kkado
and developed with
@jtleek
and
@fredhutch
(feel free to retweet)
This class is meant for everyone! It is free to audit and doesn't require CS background.
Barack Obama writes: "I still like writing things out in longhand, finding that a computer gives even my roughest drafts too smooth a gloss."
This is 1000x more true in LaTeX, where anything I write looks like a proof of the Riemann hypothesis regardless of how silly it is.
Fwiw, I plan to mask inside/avoid most crowded indoor activities because
1. Vaccines are meh against long COVID
2. Long COVID is not that rare
3. Long COVID is quite bad. I need my brain.
4. Yes, infection is inevitable, but repeated infections are possible + worse.
1/2
Moving to NYC today - as my train arrived I put on Sinatra’s “New York, New York” but then felt self-conscious cuz I didn’t think sophisticated New Yorkers would think it was very cool…
But then I went in to the subway and the subway musician was playing it too! Excellent city!
New working paper quantifying arXiv publication patterns in the age of LLMs! Joint work with
@rajivmovva
,
@sidhikab1
,
@kennylpeng
,
@gsagostini
, and
@NikhGarg
.
We analyze LLM citation patterns, fastest growing topics, many other things. Some of our findings: 1/N
Our paper "Daily, weekly, seasonal and menstrual cycles in women’s mood, behaviour and vital signs" is out today in
@NatureHumBehav
(). This is joint work with fantastic collaborators
@timalthoff
, Daniel Thomas,
@hillarpa
, and
@jure
. 1/
85% of LLM papers studying equity impacts focus on equity *harms*
This is a vital discussion. But equally vital is the more opportunity-focused counterpoint: “what new *opportunities* do LLMs enable that could promote equity?”
Our new preprint presents 4 opportunities. 1/3
Hello! I am looking for DATA SCIENCE PAPERS WHICH RESULTED IN (positive) REAL WORLD CHANGE
eg:
Nominate your favorites!
(data science here is broadly defined)
(feel free to reshare)
A few years after I completed SLE, when I was 21 or so, they asked me to give a speech at the SLE banquet, and I thought I'd post that as a tribute to Mancall and what he built. Later, I adapted this into my essay for graduate fellowships, and I believe it still.
Conversely, when I see a faculty mentor students well, I make it a point to tell as many people as possible - for example, I include them as an example when I give talks on mentorship.
Academia under-rewards faculty for treating people decently. We can help shift that balance.
I am pretty worried about the lack of demographic diversity in this list.
The team's stated goal is to produce alignment with "human intent". But "human intent" varies dramatically by the human! There are profound, well-documented demographic schisms 1/2
My paper, "Assessing racial inequality in COVID-19 testing with Bayesian threshold tests" (), won the "Best on Theme" paper award at the NeurIPS Machine Learning for Healthcare workshop (theme being "Advancing Healthcare for All").
One of the first great pieces of research advice I got, from a faculty mentor, was to bang your hand on the desk really loudly and yell BAM!!!!! when you made some research progress, because external recognition is rare. A decade later I'm still doing it.
Reading some math - I think it would be good, in mathematical writing, to minimize use of words like “trivially”, “clearly”, and “obviously” because 1) it’s unnecessary verbiage and 2) it makes the reader feel bad if they don’t immediately see what you’re saying they should.
Honored to have been named an AI2050 fellow by Schmidt Sciences. This award recognizes work with a wonderful group of students and collaborators, and I'm very much looking forward to working with the other fellows.
this is interesting!! Relatedly,
@_KarenHao
, we examined papers about large language models and documented a pronounced US/China schism - American/Chinese institutions collaborated rarely (Microsoft was an exception to this rule).
For decades the US & China have set aside differences to collaborate on science. The partnership has stabilized relations & formed the bedrock of modern research.
Now some Republicans are pushing to end it. That could be bad for the US & the world. 🧵 1/
I wanted to give a shoutout to senior women at Cornell -
@deborahestrin
,
@tanzeemc
, Eva Tardos,
@wendyju
,
@nicki_dell
,
@karen_ec_levy
, and others - for doing lion's work to help me find my feet as junior faculty. I had a good feeling about joining a school with many senior women!
Hello hello! Having spent less than 1 week living on
@cornell_tech
's New York City campus I must STRONGLY and SINCERELY recommend to PhD applicants that you consider doing your PhD here (videos are timelapses out my window).
It is beautiful in good weather...1/
I worried that, after I left my PhD lab, I wouldn't have enough ideas to support my own research lab. This turned out to be a dumb worry: I have tons of ideas. The REAL problem is lots of them are bad :P So I'm curious: what is your process for determining which ideas to pursue?
For my last lecture in my class I tried to answer the question students most often ask: "how do I learn to use data for social good?" Here are my slides which provide learning resources, places to work and broader life advice. Feedback welcome.
Our COVID-19 paper is out in
@nature
! Working with this fantastic team was an unexpected blessing during a tough year. If you write that much code, run that many models, and listen to that much
@taylorswift13
during a difficult and lonely time, you end up as friends.
At the end of a colorless winter day we got a sunset so spectacular that I panicked and ran around my apartment trying to capture it. You really never can but this was the closest I got.
On Monday I had no idea
On Tuesday I had an idea I was excited about
On Wednesday I realized my idea was flawed
Now I am in a worse mood than I was on Monday! What! That doesn't make sense. I'm right back in the same state.
#ResearchIsHard
I have an article in the New England Journal of Medicine
@NEJM
about accuracy and equity in risk prediction, based on my own experiences as a patient. Thanks to colleagues, family, peer reviewers, and editors for very helpful comments.
@DorsaSadigh
,
@percyliang
, and
@PangWeiKoh
were nice enough to invite me to give a guest lecture in Stanford's AI class, and I concluded by trying to persuade students to make the world a better place (see attached text) (1/4).
Man, academia for me is such a bittersweet combination of 1) wonderful ideas, talks, mentors, and collaborators and 2) random sexist BS. I really hope 1) will keep making it all worth it, but the bad actors so demoralize me that some days I just get real tired.
On the rare occasions when I try to collect my own data from actual people, I'm struck by how hard it is, how every tiny detail of wording matters. Hats off to folks who do this for a living.
this is well-said. In general, I take critiques of LLMs much more seriously if they start by acknowledging the obvious - these models, though flawed, are also pretty astonishing.
There’s one existential risk I’m certain LLMs pose and that’s to the credibility of the field of FAccT / Ethical AI if we keep pushing the snake oil narrative about them.
I am drafting a piece about sexism in the sciences and I have written the first paragraph four times, from scratch, to try to sound as not-angry as possible. Please keep in mind that by the time a woman expresses her frustration to you, you're often reading the fourth draft.
Since it's offer season for PhD and faculty applicants: I feel very grateful and lucky to get to work at
@cornell_tech
. (It helps that it's 60 degrees and sunny in NYC today.) If you get a chance to work here, please consider it, and come visit so you can see what I mean!
I wrote a Python script yesterday to print out your Google calendar availability to make it faster to write scheduling emails, some folks asked for it, it's online now -
this is just a vacation hack so use with a bit of care and lmk if you find bugs :)
I've been thinking about mental health/quality of life among PhD students recently. If you are or were a PhD student, please help me out by answering this short survey :)
In case it comforts nervous folks on the job market, I will say that if I'm interviewing you or reading your resume, I'm rooting for you to succeed, not trying to make you fail - it's fun to find great researchers.
It was an honor to give the graduation keynote at Thomas Jefferson High School for Science and Technology
@TJHSST_Official
. Congratulations to all the graduates - TJ is a special place, and I will always feel very lucky to have gone there.
We have decided to name our kitten sisters Mew and Sigma, for the mean and standard deviation of a normal distribution. Mew is the blue collared cat perched quietly in one spot; Sigma is the one spread out all over the bed, and is more erratic in general.
Oh goodness I just learned I might get kittens...and they were born on Easter Sunday...which is the same day I got my first vaccine shot...
Like that's got to be a good omen, right? We can call them Pfizer and Moderna. Or Johnson and Johnson (but that seems confusing).
@JeffDean
has spent several hours today reading and retweeting stories from women in tech about unwanted sexual advances. It's rare and refreshing to see a powerful and busy man spending his time this way - really made my day and would make me more eager to work at Google Brain.
Just spent an hour staring out the window and still can’t solve my math problem, going to play StarCraft and hope the answer will just come to me. That’s how that works, right?
Am proud to report that I am achieving at least 3% of my normal productivity, courtesy of the 7 neurons that showed up to duty this morning. Good job guys! The other ~10^11 will stop checking election results soon.
CS researchers: have you ever used Grammarly to proofread conference papers (ie, those written in LaTeX?) How do you use it, if so? If not, are there other tools/approaches you use? Feel free to retweet this question - interested in a diverse sample.
So yesterday I asked you all what you wanted to hear about from me this week, and one answer stood out from all the others: the SINGULAR VALUE DECOMPOSITION (SVD).
1/
We're releasing a new dataset (feel free to reshare)! 2.4 million dashcam images from NYC over a few months. High spatiotemporal density, useful for work in the social sciences/urban studies/computer vision.
Great work from
@mattwfranchi
@wendyju
to anonymize+release this data!
We’re releasing a new research dataset of 2.4 million street-level dashcam images, spanning all five boroughs of New York City from March-November 2020.
It's been a grim year, so I made a list of my favorite academic moments to remind myself of all the good things in the world: . What are yours?
Also, I wrote a song which I include in the post! Let me know if you can sing! We can be co-first authors.
I've been trying so hard to make StarCraft platinum league
With my rating at the threshold, I messaged my opponent "will you resign so I can be platinum league"
and he did!
AND THEN I WAS NOT PROMOTED
I've written 4 papers on threshold rules how did I misunderstand this one?!
I was recently asked why it's worth doing studies which prove discrimination is real. The people facing discrimination - the very people these studies are claiming to help - already know it's real!
It's a thought-provoking question. Here's my answer:
Research tip: go outside at least once a day, even if you work from home.
Every time I step outside and see this cluster of four trees - they sparkle if there’s any sunlight at all - my mood instantly lifts.
In the last three days my 87-year old grandmother has a) debugged a wifi connection issue which I could not solve b) watched my talk and told me I "need to talk slower" and c) ordered me to take more vacation.
I'll close by mentioning my conflict of interest: Pang Wei will be on the job market next year and I want an American institution to hire him because it'll be a lot more annoying to collaborate with him if he goes back to Singapore. Please hire him. Thank you.
My bf just took hours out of his workday to get forms from the DMV, drive home, get my signature, and drive back, all so I can watch a talk. This is feminism, man. We don't need a bunch of empty rhetoric about Simone de Beauvoir or whatnot. We need you to get the damn forms.
Spring at Princeton is absolutely stunning :) A pleasure to speak at the Quantitative Social Science Colloquium - many thanks to
@PUPolitics
for the invitation!
Going for a long walk in fresh snow is a great metaphor for writing a paper. On the way out you’re like “awesome! So exciting! Breaking new ground!” And by the end of it you’re like “Jesus this is exhausting why on earth did I take such an ambitious route”.
Are you more likely to meet somebody different from you in a big city or in a small town?
Check out our new paper in
@Nature
. We refute the myth that cities are cosmopolitan melting pots, finding extreme segregation in big cities.
See Ziad's summary of our just-published
@NatureMedicine
paper. This paper had the highest professor:PhD student ratio of anything I've worked on - felt very lucky because all 4 professors (
@Cutler_econ
@jure
@m_sendhil
@oziadias
) were great mentors.
Medical puzzle:
Take Black vs. White patients with similar knee x-rays
Why do Black patients have more pain?
Our paper in
@NatureMedicine
provides one answer:
Algorithms see causes of knee pain in Black patients, that human radiologists miss.
🧵
Post-doc opening
@cornell_tech
! In "urban tech" but this is broadly defined - can work with whomever you want, centrally funded.
Happy to chat if you're interested in inequality/health/ML and their intersections with city life :) Feel free to reshare.
I am giving a review talk on algorithmic fairness tomorrow, and no author is cited more than
@timnitGebru
. Her research is groundbreaking. Her leadership equally so. I do not know the details of what happened here, but I do know our community has benefited hugely from her work.
Tension I struggle with: most great scientists I know are meticulous - eagle-eyed for flaws in an analysis, weaknesses in an argument. But, looking at work this critically can demoralize you and others. How do you balance being self-critical while still taking joy in science?
Research tip: make large quantities of soup and stew and freeze it. My grandmother did this when finishing her PhD. I make terabytes of minestrone, freeze it, and heat it up with frozen sourdough rolls so I have hot soup and hot bread when I’m too tired to cook.