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.
Today we’re launching AI21 Studio, a platform where you can instantly access our state-of-the-art language models to build your own applications - including a 178B parameters model, Jurassic-1 Jumbo. We can’t wait to see what you create!
https://t.co/NhCnF1Emcy
We think the spotlight can fit into practitioners' toolkits, fitting into a broader feedback loop of developing, auditing, and repairing models.
Collaboration with @jasondeon_ and my co-advisors James Wright, @k_leyton_brown.
My first PhD paper is up! We develop "the spotlight", a method for identifying systematic errors in trained deep learning models without access to side information like protected features.
Paper: https://t.co/HfFE0T1AzX
Code: https://t.co/qKRFt9IoW5
...an Amazon review classifier misunderstands Spanish reviews, and a matrix completion model is often overconfident about reviews for Pulp Fiction! (More datasets and spotlights in the full paper.)