🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

On Saturday, I will be giving a talk at 11:40 am at the ICML @agenticmarkets workshop on our work on three-layer data markets: arxiv.org/abs/2402.09697. I would discuss a model to study data monetization and the role of privacy regulations in the presence of strategic agents!

This year, I have 4 exceptional students on the academic job market, and they couldn’t be more diffferent, with research spanning AI policy, robotics, NLP, and HCI. Here’s a brief summary of their research, along with one representative work each:

Probably should be making pies, but its much more fun to blog about calibration. This one gets into more practical issues w/recently proposed solutions like multicalibration, discussing results from a recent empirical evaluation statmodeling.stat.columbia.edu/2024/11/26/pra…

1/ Thrilled to share an advancement in gene therapy from my PhD in Nature Communications! We've developed a new approach to reduce immune responses while maintaining efficiency—paving the way for safer, more effective therapies. Big thanks to Feng Zhang & Mirco J. Friedrich! bit.ly/redicas9

This morning, the Joint California Policy Working Group on AI Frontier Models released its draft report. At the request of Governor Newsom, the report outlines guiding policy principles that can inform California's frontier AI policy. A quick 🧵:

Want state-of-the-art data curation, data poisoning & more? Just do gradient descent! w/ Andrew Ilyas Ben Chen Axel Feldmann Billy Moses Aleksander Madry: we show how to optimize final model loss wrt any continuous variable. Key idea: Metagradients (grads through model training)

Had a great time at Simons Institute for the Theory of Computing talking about new & upcoming work on meta-optimization of ML training tl;dr we show how to compute gradients *through* the training process & use them to optimize training. Immediate big gains on data selection, poisoning, attribution & more!

This work has been a long time coming and I'm so grateful to my collaborators for helping make this work possible. Main takeaway: AI supply chains matter! We've seen them emerge (rapidly) in the past few years and they will have implications on *all* of us, inside and outside

We’re organizing a tutorial on the Economics of Data as part of ACM SIGecom EC ’25! 🗓️ June 18 | 🕑 2–4 PM ET | 📍Zoom Registration is free but mandatory! Details and registration link are available on the tutorial webpage!👇

Today, Co-Leads of the Joint CA Policy Working Group submitted the California Report on Frontier AI Policy to @CAGovernor’s office. Informed by public and expert feedback, this final report outlines a policy framework for frontier AI development. Read more:

“How will my model behave if I change the training data?” Recent(-ish) work w/ Logan Engstrom: we nearly *perfectly* predict ML model behavior as a function of training data, saturating benchmarks for this problem (called “data attribution”).