Congzheng Song

Privacy research @OpenAI | Ex-Apple | PhD CS from Cornell Tech

[Congzheng Song @csong27]

RT @awnihannun: If you're into federated learning / private ML, check-out Apple's PFL reasearch framework by @Grananqvist and team.

It's a…

[Granqvist @Grananqvist]

I'm excited to announce that we have open sourced pfl-research — a fast, modular and easytouse Python framework for simulating Private Federated Learning for research.
A great milestone toward reproducible research in PFL. Give it a⭐️🙂 https://t.co/1RpSF2UplP #FederatedLearning

[Eugene Bagdasarian @ebagdasa]

I am looking for PhD students to work together on privacy and security problems in “AI Systems”. We will focus on language models, agents, ML services, and study where they fail and how to make them work better. Apply by December 15. #phdlife #manningcics

[Eugene Bagdasarian @ebagdasa]

NEWS: I am on the academic market this year (Privacy and Security, Trustworthy AI). At @cornell_tech I studied how ML-based applications and systems can fail and/or cause harm and how to make them better. My package is here: https://t.co/o8aFSCSdcl . Happy to chat at #NeurIPS2022

[Congzheng Song @csong27]

@pag_crypto Congrats Paul!

[Reza Shokri @rzshokri]

I am hiring PhD students who are interested in any of the following areas: data privacy, algorithmic fairness, algorithmic transparency. Apply now to CS dept. Prateek Saxena @prateekatcs and I have open positions for postdocs in data protection for decentralized computation.

[Michigan Systems Lab @MichiganSystems]

Starting this fall at @UMichCSE, meet @pag_crypto! His research at the intersection of cryptography and systems has already had broad impacts across the IT industry. Look forward to seeing what he'll accomplish at U-M!

https://t.co/rrLrAPbGMU

[Reza Shokri @rzshokri]

When designing trustworthy machine learning algorithms, it is very important to have a clear understanding of the relation and trade-offs between different requirements. We analyze the privacy risks of providing group fairness for ML. https://t.co/Bad64UCIm4 We show how

[Florian Tramèr @florian_tramer]

Interested in looking for memorized training data in GPT-2 or other LMs?

I've released some code to get started here: https://t.co/QfxjJ0lDrT

And a list of ~50 weird things we extracted from GPT-2 here: https://t.co/GJsKFyxigl

If you find other fun stuff, please let me know!

[Congzheng Song @csong27]

@YVigfusson @SimbiosysLab @emory_csi @Rebecca4Georgia Congrats Ymir!

[Congzheng Song @csong27]

@rzshokri @IEEESSP @JPHubaux Congrats Reza!

[Reza Shokri @rzshokri]

We analyze the dynamics of privacy loss to address a major challenge in differential privacy for iterative ML algorithms: quantifying the privacy loss of the “released” model, when internal state of the algorithm (i.e., intermediate params) is private. https://t.co/R66Dh1cnuI

[NAFSA @NAFSA]

NAFSA continues to engage with @DHSgov and members of Congress on the severe #f1optlockboxdelay impacting #intlstudents and to push @USCIS to announce solutions ASAP. Retweet the following recommendations to show your support! #intled

[NAFSA @NAFSA]

NAFSA is proud to be among the signers of @ACEducation-led letter to @DHSgov RE: the importance of supporting #intlstudents & #OPT applicants' need for flexibility during this unprecedented time. Monitor our site for updates. #intled #f1optlockboxdelay https://t.co/hHsYtLEf6E

[Ymir Vigfusson @YVigfusson]

The billing metadata collected by all cell-phone providers around the world could prove useful for public health and outbreak surveillance. Our PNAS paper shows infection-associated behavioral changes in data from Iceland.
https://t.co/tuIBA7kcOU @PNASNews #epitwitter

[Rebecca Mitchell, DVM, PhD @RebeccaUpstate]

The first, and only, PNAS paper this state legislator will be on. @314action

Sometimes you make the policy, and sometimes you help with the research that makes the policy.

https://t.co/3wgk0yjNoC

[Leon @leondanon]

Cell-phone traces reveal infection-associated behavioral change https://t.co/uZQjxjpsPi

[Reza Shokri @rzshokri]

The main objective of paper review process is not paper selection. Spoiler alert: It is rather providing an objective technical review of a paper from the viewpoint of an expert. The discussion between reviewers should also be primarily about this. The decision making process

[Behnam Neyshabur @bneyshabur]

Some people say that one shouldn't care about publication and the quality matters. However, the job market punishes those who don’t have publications in top ML venues. I empathize with students and newcomers to ML whose good papers are not getting accepted. #ICLR2021
1/

[Gautam Kamath @thegautamkamath]

Inspired by Ryan (@BooleanAnalysis), I made a Bilibili account, and started uploading my lectures & talks. The eager audience there is too large to ignore.

I naturally started with my course, but there'll be more to come. Be sure to 素质三连走一发，谢谢!
https://t.co/Yy4cEV4Hon