Come hear about work on datamodels (arxiv.org/abs/2202.00622) at ICML *tomorrow* in the Deep Learning/Optimization track (Rm 309)! The presentation is at 4:50 with a poster session at 6:30. Joint work with Sam Park Logan Engstrom Guillaume Leclerc Aleksander Madry

Was great talking to the amazing Brazilian data management community SBBD 2022 about our Knowledge Graph @apple more details machinelearning.apple.com/research/conti…

Big news from Formerly Axelar, now at @axelar congratulations on the launch!

Congratulations Jimmy Jimmy Lin

Based on the latest groundbreaking #GPT4 model from OpenAI, we had the opportunity to experiment and learn more about its capabilities. I would not have believed just six months ago that I am now saying that there are sparks of #AGI in the model. Amazing times.

As ML models/datasets get bigger + more opaque, we need a *scalable* way to ask: where in the *data* did a prediction come from? Presenting TRAK: data attribution with (significantly) better speed/efficacy tradeoffs: w/ Sam Park Kristian Georgiev Andrew Ilyas Guillaume Leclerc 1/6

The future of AI is many smaller "specialist" models that are faster *and* cheaper on specific datasets & use cases- not one large "generalist" model. Very excited to share our work on doing this via more efficient distillation, with Google AI & led by the amazing Cheng-Yu Hsieh !!

First Egyptian and second African to the quarterfinals #ITTFWorlds2023

"Data Science–A Systematic Treatment," by M. Tamer @Ozsu University of Waterloo, discusses the scope of #DataScience and highlights aspects that are not part of AI bit.ly/3JqNp5z. Özsu talks about leveraging data and deploying data science in a #video. bit.ly/3NJaToY

What makes ML models tick? How do we attribute model behavior to the training data, algorithm, architecture, or scale used in training? Papers (or ideas) here? Submit to ATTRIB @ NeurIPS 2023 (attrib-workshop.cc)! Deadline is September 23!

The deadline to submit to ATTRIB 2023 is TODAY anywhere on earth! Submission instructions here: attrib-workshop.cc

🥳 Super excited at multiple #EMNLP2023 acceptance. Two by my summer interns Apple: Simone Conia on growing multilingual KG and Farima Fatahi Bayat on factual error detection & correction. One with my former colleagues IBM Research on a novel labeling tool. Details coming soon.

Amazing work by team on the role of #KnowledgeGraphs in effective deployment of AI. Congrats team!

Two more ACM fellows for University of Waterloo . Congrats to my colleagues Shai and Ian. Well deserved! Association for Computing Machinery Waterloo Mathematics Waterloo's Cheriton School of Computer Science

Grateful to my students, colleagues and talented collaborators.

ConvKGYarn🧶: Spinning Configurable & Scalable Conversational KGQA datasets with LLMs is accepted at #EMNLP2024 Industry Track with a top-20% rec⭐️ Work done Apple with Daniel Lee , Ali, Jeff, Yisi Sang , Jimmy Lin , Ihab Ilyas , Saloni Potdar , Mostafa, Yunyao Li ✈️ ACL'2025!

Congrats Andrew! Can’t wait to see what you are going to do next :)

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)

“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”).