Kristian Lum @KLdivergence Twitter profile

Last Seen Profiles

@himarycake

@BrianJJonesUK

@biomejs

@DarkUnicornsCoC

@siema161

@femboygoetia

@billyatterbury

@FloraAcetyra

@Tzrri

@_DJDON

@SchweitzerTracy

@TheApacheStruts

@_tsukuyxmi

@UtahTennis

@NFX

@Chahali

@SavvyAuntie

@Classicmariposa

@laylah_lawson

@APLCRICKETUSA

@askiisoft

@IsaiahADixon

@Havas

@tunamayoyo

@Lawrenc_lin

@DennysCanada

@DHeidtke21

@yonkoujean

@5ANGE_GANK

@mura_buta

@_masoncrowe

@Suvyshotdamovie

@Benoson_1

@ChrisSteiner20

@jandakembangstw

@Unit_Shawn

Kristian Lum

@KLdivergence

6 years

I'll probably end up remembering this moment as the time I decided to blow my career up for no reason, but this is something I've been wanting to say for a long time.

Statistics, we have a problem.

Recently, while browsing Twitter, I saw a few machine learning researchers post about an incident at one of their big conferences (NIPS) in…

medium.com

374

3K

6K

Kristian Lum

@KLdivergence

2 years

🫡 and I’m out. Condolences to the survivors. #lovewhereyouworked

76

84

2K

Kristian Lum

@KLdivergence

4 years

I’ve got a statistics joke but it will be way more marketable if I call it an AI joke.

21

170

2K

Kristian Lum

@KLdivergence

3 years

Literally a review I just got on a paper I submitted: "This could have been one of the shortest reviews I have ever written: I think this paper is excellent and I cannot think of anything I would suggest to improve it." Miracles do happen. Thank you, Review 1. Thank you. 🙏

12

21

2K

Kristian Lum

@KLdivergence

1 year

First day at the new job vibes.

68

27

2K

Kristian Lum

@KLdivergence

3 years

No matter how far you've made it in academia, there's no shame in deciding it's all bullshit and just.... doing something else.

49

84

1K

Kristian Lum

@KLdivergence

3 years

Woman, not girl. And good to know credentials are important to you because I have a PhD in Statistics. Enjoy the video!

23

24

1K

Kristian Lum

@KLdivergence

4 years

Change my mind: the biggest difference between statistics and machine learning is that statisticians think a lot about sampling and people in ML largely...don't.

43

103

995

Kristian Lum

@KLdivergence

3 years

People with young children who work full time: how do you find time to exercise? It seems like it comes down to picking between sleep and exercise (faced with that decision, I'll always pick sleep). Tips actually, genuinely appreciated!

426

30

964

Kristian Lum

@KLdivergence

2 years

All of twitter’s ML Ethics, transparency, and accountability team (except one). was laid off today. So much for that responsible ML initiative, huh?

27

147

871

Kristian Lum

@KLdivergence

1 year

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.

41

139

865

Kristian Lum

@KLdivergence

1 year

I’d always assumed layoffs were stressful for people because of financial concerns. I’m not even that worried about money right now, but the feeling of being discarded, tying self-worth and identity to my job, loss of mission… all of that is harder than I expected. That’s all.

37

39

850

Kristian Lum

@KLdivergence

6 years

Just got a message from a HIGH SCHOOL STUDENT who had pretty much reproduced one of my published papers... for a class project. Kids these days, man.

11

53

707

Kristian Lum

@KLdivergence

4 years

Proud to announce that for the 35th year straight, I got 0 papers accepted out of 0 submitted at #NeurIPS2020 ! Wonder if I can pull it off again next year!?

13

17

668

Kristian Lum

@KLdivergence

3 years

For a while, I've been thinking about making YouTube videos about stats/data analysis, etc. Here's my first one! It's about debunking some recent analyses that claim to prove that there was election rigging. Tl;dw there was not.

35

115

650

Kristian Lum

@KLdivergence

4 years

Some exciting news: today I joined the CIS Department at @Penn as research faculty!

44

10

640

Kristian Lum

@KLdivergence

6 years

Seeing my essay cited as a reason that prominent people in statistics and ML are taking a stand for a more inclusive conference culture makes the stress of the last several months all worth it.

Dr. Elana J Fertig

@FertigLab

6 years

I'm proud to be a member of the @JohnsHopkins faculty making a call for greater inclusion for #MachineLearning and #WomenInSTEM at @NipsConference

10

81

273

11

71

629

Kristian Lum

@KLdivergence

18 days

I will never get over how AI/ML people use the word “inference”

67

35

606

Kristian Lum

@KLdivergence

3 years

I really don’t get LinkedIn. About once a month I log in and just accept all of the invitations to connect…. And that’s basically it. Are there people who actually use it for networking??

82

10

589

Kristian Lum

@KLdivergence

3 years

A dude who once told me he was "kind of a famous data scientist"-- presumably based on his Twitter follower count. Just stumbled across him on here & realized I now have more followers than him.... and I still cannot imagine those words ever leaving my mouth.

31

8

575

Kristian Lum

@KLdivergence

4 years

I spent 2 years building realistic simulated contact networks to run epidemic (& other) simulations. Much of my recent work is on estimating mortality undercounts. If there are any covid researchers out there who could use a few more (well-washed) hands on deck, lmk.

24

148

565

Kristian Lum

@KLdivergence

4 years

I honestly don't understand how some people crank out so many papers a year. It takes me about a year to produce one piece of work that I'm proud of. And it's so complete that small extension papers seem pointless. What I'm saying is-- I need to learn how to play the game better

45

23

531

Kristian Lum

@KLdivergence

3 years

Making some FAccT / ethical AI / algorithmic fairness meeting bingo cards. What should I put on them? So far I've got...

59

99

525

Kristian Lum

@KLdivergence

2 years

So turns out doing an interview a few hours after getting laid off is not a great idea…

28

1

479

Kristian Lum

@KLdivergence

1 year

Proposal to study the effect of h1b visas on making people stay in toxic jobs: a case study on “the fork in the road” at Twitter.

12

37

470

Kristian Lum

@KLdivergence

3 years

With the topic of PhD women using the Dr honorific back in the news, time for one of the most bizarre interactions I’ve had related to this. Them: Ms. lum, please speak at our workshop! Me: sure! Btw it’s Dr. Lum but first name is fine. Them: oh I’m so sorry, sir. Me: 🤦🏻‍♀️

11

13

432

Kristian Lum

@KLdivergence

3 years

On a plane and the people in the row in front of me have so far said “degrees of freedom”, “rank”, and “column space” And I want to talk to them so badly.

8

1

404

Kristian Lum

@KLdivergence

4 years

Hi @Spotify . Every time I put on kids music it ends up auto playing to Christian kids songs. Can you, at the very least, let me opt out of the religious indoctrination music? I’m trying to raise a good atheist here. Thanks.

13

9

381

Kristian Lum

@KLdivergence

2 years

So many emotions about the lay off. Relief because it’s all over. Gratitude for having gotten to work with the brightest, most compassionate people around. Pride for what we did accomplish. And sadness for what we could and would have accomplished in a different timeline.

20

21

369

Kristian Lum

@KLdivergence

4 years

PhD at 25 here. Totally agree with this. I would have gotten much more out of grad school with a little more real world experience first.

Dr. Tao Leigh Goffe | 道 🐲🌋

@taoleighgoffe

4 years

I finished my PhD at 26. I regret this. Take time off between grad school and undergrad. Live your life. Defer, read, work, explore, party, save, make art, sleep. Don’t rush. Grad school is not the safety net you think it is. The degree will be there when you’re done. Enjoy life!

2K

49K

307K

21

18

361

Kristian Lum

@KLdivergence

4 years

Seriously, how is anyone getting anything done right now? Every time I try I end up just anxiously reading and re-reading covid news. Tips welcome!

22

14

354

Kristian Lum

@KLdivergence

4 years

For a person who gets constantly critiqued for the way I give talks (too fast, too informal, etc), I get an awful lot of speaking invitations. 💅🏼😘

6

3

342

Kristian Lum

@KLdivergence

4 years

Had the amazing experience of having someone contribute to a #RStats package of mine.... and now it's running 75x faster! What a highlight to know that someone liked what we did enough to improve on it!

4

9

330

Kristian Lum

@KLdivergence

3 years

I've seen FEAT and FATE as acronyms for (obviously different orderings of) Fairness, Accountability, Transparency, and Ethics. But, why has no one chosen FETA? Much more delicious than the other options.

19

16

325

Kristian Lum

@KLdivergence

2 years

This is disturbing, but not for the reason many people are implying. These are the faces that *humans*-- not AI-- rated as most/least trustworthy.

\newcommand{\femb0t}{

@__femb0t

2 years

Src : BRUH

45

34

589

22

56

317

Kristian Lum

@KLdivergence

3 years

If your first reaction to hearing that someone willingly left an academic job in your field is to imply to them it’s because they couldn’t hack it, maybe it’s not so much of a profession as it is a cult.

11

13

321

Kristian Lum

@KLdivergence

2 months

It finally happened. Someone asked me a question and the answer was something in my dissertation. I’ve been waiting for this moment for almost 15 years.

6

11

321

Kristian Lum

@KLdivergence

6 years

This week has been insane. Now back to work, so I can hopefully be known for something other than sexual harassment whistleblowing.

4

2

312

Kristian Lum

@KLdivergence

2 years

Evaluating subgroup disparities in ML shouldn't only be a "fairness" thing. It should be ML/data science best practice in general. If your model performs poorly for some identifiable subgroup, you should probably go fix that, just in the interest of making a better model.

16

49

305

Kristian Lum

@KLdivergence

1 year

🫡 to all the brand new Twitter alumni.

2

14

301

Kristian Lum

@KLdivergence

5 years

Roses are red ML is fun But please pay attention To where your data comes from

Christopher D. Long

@octonion

5 years

Roses are red Violets are blue Small data for me Is big data for you

0

4

40

2

75

302

Kristian Lum

@KLdivergence

3 years

My unsolicited advice to academic job seekers: if there is a department that keeps losing faculty, don't believe that the departures are random. The place is a shit show; run like hell.

10

18

301

Kristian Lum

@KLdivergence

1 year

This isn’t to say that they don’t make stupid mistakes, hallucinate, or “just” memorize training data. They do; I'm convinced. But the idea that this negates what is obvious to anyone who has signed up for a ChatGPT account— that they are incredibly powerful — is absurd.

10

34

302

Kristian Lum

@KLdivergence

4 years

I might as well just go ahead and bookmark the google search results for "change ggplot legend title", because after **years** of doing this, I still have to look it up every.single.time.

13

290

Kristian Lum

@KLdivergence

3 years

I've always been drawn to thorny problems that require quantitative sophistication coupled with a compassionate understanding of complex social issues. And, oh my, is tech full of those kinds of problems! So, 🌟 big announcement 🌟.... 🥁🥁🥁 It's true! I'm joining Twitter!

18

16

293

Kristian Lum

@KLdivergence

3 years

It's great to see mainstream acknowledgment that studying AI impacts is both important and interesting. The questions below are exactly the questions I've been highlighting and working on for years, & this work has consistently gotten me labelled as "not technical enough."

6

53

292

Kristian Lum

@KLdivergence

3 years

@math_rachel While I disagree with the perspective that a solution must be offered in order to point out a problem, it’s worth noting that work in the AI ethics space often does point to a solution..... don’t build it. Or change policy. These often don’t count because they aren’t “technical”.

5

53

291

Kristian Lum

@KLdivergence

2 years

If I stand by an nba locker room and record the heights of people going in, that is not a representative sample of the population of the US. This remains True even if I record every nth person.

9

16

292

Kristian Lum

@KLdivergence

3 years

I love, and I mean this genuinely, really LOVE doing data analysis and building models. At the same time, spending a large part of my life staring into screens feels like a real waste. I think this is going to be a conflict that's hard to resolve for me.

18

8

283

Kristian Lum

@KLdivergence

5 years

You know you're doing alright when you start taking things *off* your CV.

2

10

272

Kristian Lum

@KLdivergence

3 years

LOL that feeling when the editor who just rejected your paper follows you on Twitter while you're complaining about it. Hi there 🤚

4

3

263

Kristian Lum

@KLdivergence

3 years

2 final important things: 1) fuck you, Walter lasecki and everyone who protected him. 2) I sadly do have experience being at the center of very public sexual harassment/assault debacles. If anyone needs support from someone who’s been through it, don’t hesitate to reach out.

2

12

262

Kristian Lum

@KLdivergence

4 years

Subtweet incoming: No matter how righteous you consider your crusade, if you don’t frequently stop to seriously consider the possibility that you’re actually the bad guy, I’ve got bad news for you..... you’re probably the bad guy.

9

32

252

Kristian Lum

@KLdivergence

3 years

Ok, so I get I'm way behind the times here, but wtaffffffff is this probabilistic machine learning business? How is this different than statistical/stochastic modeling?

28

21

245

Kristian Lum

@KLdivergence

2 years

I love Twitter. I love coming here to learn about new research, see the latest news, connect with people, and get the hottest takes. I know this hasn't been everyone's experience, but for me, this has been a magical place where I've found community and really felt heard.

4

9

248

Kristian Lum

@KLdivergence

3 years

Do you use the COMPAS dataset to demonstrate fair ML methods? Do you want to do so responsibly? Then, "It's COMPASlicated," my new paper w @TheMichelleBao @angelamczhou @samzottola @BrianBrubach @DrSLDesmarais @Aaron_Horowitz @geomblog is for you! 🧵 1/n

8

73

248

Kristian Lum

@KLdivergence

4 years

Academic publishing can be completely demoralizing. Just this week: -on 4th revision, asked to undo work that was explicitly requested in last round -paper rejected after months-long wait based on 1 detailed, extremely positive review & 1 unpunctuated, 125 word nonsensical review

20

15

240

Kristian Lum

@KLdivergence

5 years

Getting home from #babysfirstacademicconference and working/traveling with an infant was.... not so bad. I’ve got this.

6

3

239

Kristian Lum

@KLdivergence

5 years

STOP talking shit about different presentation software PowerPoint is CLASSIC Keynote is CLEAN Prezi is INNOVATIVE Beamer Google Slides is CONVENIENT

22

27

236

Kristian Lum

@KLdivergence

3 years

You know that feeling when you have a really good research idea that seems so good it's impossible that no one's thought of it before... and yet, you can't seem to find it anywhere? It's kind of a combination of excitement and fear.

10

4

236

Kristian Lum

@KLdivergence

4 years

@AmihaiGlazer Lol what’s the R-squared on this bad boy?

4

3

230

Kristian Lum

@KLdivergence

2 years

Due to a family emergency it looks like I’ll be flying solo at #JSM2022 with my 3 year old in tow. Normally I’d just skip it but this year is special because for the first time in my career I’m getting an award. Hoping she allows me to walk up and get it without making a scene.

22

3

233

Kristian Lum

@KLdivergence

5 years

Excited to announce I'll be moving to Philadelphia this summer! I'll still be working at HRDAG, but I'll be looking for opportunities to find collaborators and deepen professional connections locally in Philly and NYC. Ideas and introductions very appreciated!

32

8

234

Kristian Lum

@KLdivergence

1 year

If the field is going to survive this “disruption” (sorry, cringe), we need to meet reality where it is, not hold on to a narrative that is so easily falsifiable by anyone with a smart phone and 10 minutes to spare.

9

4

233

Kristian Lum

@KLdivergence

4 years

Today I found out that a group of adults at the park was making fun of my 1.75 y/o daughter for "looking like a boy" because she has short hair. Shaming kids for not conforming to gendered beauty standards starts earlier than I'd thought. Lord help me with this parenting thing.

23

8

230

Kristian Lum

@KLdivergence

3 years

One big difference between tech and previous jobs I've noticed so far is the default to the half hour meeting. It's like a firehose, but I'm never going back. Hour-long meetings: never again.

15

3

229

Kristian Lum

@KLdivergence

4 years

Ugh, that feeling of making your paper materially worse by incorporating the suggestions of idiot reviewer #2 but knowing if you don't, they will keep you in review purgatory for endless rounds of revisions.

13

4

229

Kristian Lum

@KLdivergence

2 years

To all the people showing up in tweeps’ mentions to revel in the layoffs because we had it better than they did: us having good salaries and humane working conditions didn’t lessen your ability to have that too. We weren’t hoarding those things; we want them for you too.

7

21

225

Kristian Lum

@KLdivergence

3 years

logged in to Twitter to see that we that we are doing the whole “algorithmic bias is just a data problem and we can fix it” thing again. Getting my tweeting thumbs ready, but actually... Nope. Not today, satan. Logging out for the rest of the weekend.

7

223

Kristian Lum

@KLdivergence

2 years

Can someone please get Elon an intro stats textbook?

6

11

220

Kristian Lum

@KLdivergence

4 years

“Biased algorithms are easier to fix than biased people”. But, in some cases, I don't see these as entirely different problems. Sometimes fixing a biased algorithm first requires fixing biased people.

3

64

219

Kristian Lum

@KLdivergence

2 years

Risk assessment instruments are used in the criminal justice system to estimate 'the risk a defendant poses to society'. We developed a tool to flip the script & estimate the risk the system poses to a defendant. A 🧵 on my latest w @mmeyer717 @Aaron_Horowitz & Erica Marshall

9

59

218

Kristian Lum

@KLdivergence

2 years

Excited to tell you about my paper with @amandarg__ and @zywind on measuring model performance disparities (aka “bias”). mini-🧵

De-biasing "bias" measurement

When a model's performance differs across socially or culturally relevant groups--like race, gender, or the intersections of many such groups--it is often called "biased." While much of the work...

arxiv.org

5

43

214

Kristian Lum

@KLdivergence

3 years

I kind of love the pre-recorded online conference format because I can watch the talks on 2x speed.

9

6

211

Kristian Lum

@KLdivergence

3 months

The band is getting back together! Tomorrow, I’m joining @wsisaac and so many others I admire on @Google DeepMind’s Ethics team to work on AI evaluation. Exciting times ahead…

20

3

207

Kristian Lum

@KLdivergence

4 years

Who's got two thumb emojis and is the "researcher from the University of Pennsylvania" who teamed up with the ACLU to build an epi model incorporating covid spread in jails? 👍This gal.👍 What a dream to work with one of my favorite organizations!

ACLU study linked to jails projects coronavirus deaths double US government estimates

The ACLU model was based on data from 1,200 of the country's jails.

abcnews.go.com

5

23

206

Kristian Lum

@KLdivergence

3 years

Why am I always the LAST person to exit out of a zoom meeting? Are all of you just sitting there with your mouse hovering over the end button, or what?

15

3

208

Kristian Lum

@KLdivergence

3 years

Hey #FAccT people, has anyone looked at the Zillow Zestimate model? Pretty sure that thing has a huge influence on the prices people actually pay for houses (which would make for some interesting evaluation challenges). A "high stakes" setting I haven't seen discussed much.

19

12

210

Kristian Lum

@KLdivergence

3 years

🎆 Personal Update (not a new job) 🎇 I'm moving to Chicago! Anyone in the windy city who likes to talk stats* with a heavy dose of AI snark and some big toddler-mom energy, hit me up. I need friends. *Or anything else, really

22

2

209

Kristian Lum

@KLdivergence

6 years

I do not envy my doctor. Got some lab results and plan to go in to explain why percentiles calculated from 17 people (assuming normality ughhhhh) in 1982 probably don't mean a whole lot.

12

10

203

Kristian Lum

@KLdivergence

5 years

Seems like most of the stats world is in Pisa right now. I’d be totally jealous if I weren’t currently spending all day staring at (and feeding) this face.

9

0

202

Kristian Lum

@KLdivergence

2 years

I had just finished having one major career overhaul when I came to twitter and was so excited to settle in and do some cool work and not rethink it for a while. Facing having to figure out what I want to do all over again less than a year later… I just can’t.

13

2

198

Kristian Lum

@KLdivergence

6 years

Congrats to @shiraamitchell who put together a github page where she compares mathematical notions of fairness with unified notation and gives context for each. A huge undertaking and great resource. Modifications/additions welcome!

3

83

196

Kristian Lum

@KLdivergence

3 years

If you use scikit-learn or R packages instead of writing your own estimation procedures from scratch, don’t worry about what’s in the vaccine.

3

9

196

Kristian Lum

@KLdivergence

2 years

I really wish Twitter would give me follow recommendations based on mutual blocks/mutes. What could possibly bring us together more than our mutual unexpected annoyance at certain accounts?

7

13

196

Kristian Lum

@KLdivergence

2 years

Feels like a good day to rebuild from scratch. I'm KLdivergence @mastodon .social Mastodon. Find me there, and I'll follow back.

16

192

Kristian Lum

@KLdivergence

6 years

This is the author of that 2x2 matrix that categorizes data-related skills that's making the rounds. He's defended his work while responding **extremely graciously** to the criticisms. Just wanted to call that out bc this type of civility is all too rare on Twitter. :)

Chris Littlewood

@filtered_chris

6 years

@StuartBuck1 Hi Stuart. I’m sorry it’s come across that way; I think I could have been clearer in the plot. In the post itself, I think it is clear I’m talking about where to focus in our (Filtered’s) context, which includes our current capabilities and the marginal value of building them...

6

2

50

5

11

193

Kristian Lum

@KLdivergence

6 years

I am so grateful to @kat_heller for going on record to back me up (and of course for all of the private support I’ve gotten behind the scenes). Kat, you’re brave and you’re my hero.

Hal Daumé III

@haldaume3

6 years

wow, I guess I'm late seeing this, but so impressed at the courage of @KLdivergence and @kat_heller for taking a stand!!!

2

33

185

3

20

195

Kristian Lum

@KLdivergence

2 years

I've had a pretty non-linear career, spending time in industry, academia, and non-profits, and the rate at which I've encountered people who are mind-blowingly brilliant is about the same across sectors. I think that would surprise most academics.

5

12

195

Kristian Lum

@KLdivergence

2 years

Just realized I am NOT locked out of my Twitter computer. No, no, no, no, there must be some mistake.

16

2

193

Kristian Lum

@KLdivergence

1 year

Just dropped my Twitter laptop off to ship back and the fedex employee who accepted my shipment treated me with more compassion about the situation than anyone at Twitter 2.0 ever has throughout this whole process.

1

3

193

Kristian Lum

@KLdivergence

6 years

@RenRidesCycles @Molly__Cohen Thanks :) I was really nervous about posting this, but you're right! There is a good reason.

2

189

Kristian Lum

@KLdivergence

2 years

I dare you to write an opening sentence of a paper that does not use the word "increasingly".

18

9

192

Kristian Lum

@KLdivergence

10 months

The last year has left me feeling professionally burned and burned out to a degree I didn’t think possible. Anyway I took a vacation to Italy for a couple weeks, wore some ridiculous hats, and came back feeling *almost* human again.

7

1

192

Kristian Lum

@KLdivergence

2 years

It ain’t over til it’s over but I left a perfectly awesome academic job to come to work at twitter less than two years ago… and I don’t regret a thing. I’ve learned so much and have gotten to work with the best of the best on a platform I love. What a fucking privilege.

7

0

190

Kristian Lum

@KLdivergence

3 years

I am so glad that I applied to graduate school in stats when I did (close to 15 years ago) and not now. The level of experience and professionalism the applicants nowadays have is... well, I wouldn't have stood a chance.

9

1

191

Kristian Lum

@KLdivergence

4 years

Whoa! Didn’t know this would be in actual, physical print! Was flipping through my @AmstatNews magazine and there was my big face staring right back at me.

6

8

186

Kristian Lum

@KLdivergence

2 years

I love when I think I've figured out a really cool stats thing and then it turns out that some famous statistician figured out the same really cool stats thing like a decade (or two) before I was born.

12

2

187

Kristian Lum

@KLdivergence

2 years

Someone took a candid yet professional looking photo of me that I actually like and, in homage to my academic roots, I will now be using that photo on my website for the next 20 years.

4

2

177

Kristian Lum

@KLdivergence

3 years

If you're going to fabricate data, at least have the decency to not use a UNIFORM distribution! At least use a normal, or, like a mixture of a normal and something heavy-tailed. Sheesh!

15

8

175