
Private Online Prediction from Experts: Separations and Faster Rates
Algorithms for private online prediction from experts that improve regret bounds for non-adaptive adversaries and achieve sub-linear regret for oblivious adversaries.

Stabilizing Transformer Training by Preventing Attention Entropy Collapse
Investigating training dynamics of Transformers by tracking attention entropy to prevent attention collapse

The Monge Gap: A Regularizer to Learn All Transport Maps
The blogpost discusses the use of the Monge Gap as a regularizer in machine learning to learn transport maps efficiently.

Onboard users to Amazon SageMaker Studio with Active Directory group-specific IAM roles
Onboarding users to Amazon SageMaker Studio using Active Directory group-specific IAM roles.

Google at CVPR 2023
Google Research will showcase cutting-edge research in computer vision and machine perception at CVPR 2023, with 90 papers presented and active involvement in 40 workshops and tutorials.

YouTube: Enhancing the user experience
Utilizing AI to enhance various aspects of user experience on YouTube

Envisioning the future of computing
MIT students discuss the potential of computing advancements to reshape society in a competition hosted by the Social and Ethical Responsibilities of Computing.

SambaSafety automates custom R workload, improving driver safety with Amazon SageMaker and AWS Step Functions
Automating custom R workload at SambaSafety to improve driver safety using Amazon SageMaker and AWS Step Functions.

Symphony: Composing Interactive Interfaces for Machine Learning
The blogpost discusses the limitations of current machine learning interfaces and introduces Symphony, a new interface designed to facilitate analysis and communication between multiple stakeholders.

Cross-lingual Knowledge Transfer and Iterative Pseudo-labeling for Low-Resource Speech Recognition with Transducers
This blogpost explores cross-lingual knowledge transfer and iterative pseudo-labeling as techniques to enhance accuracy in low-resource speech recognition systems.

Black Swan Data CTO on how to tackle petabyte-level learning
Expert insights on tackling petabyte-level learning from the Black Swan Data CTO at Snorkel AI's summit in 2022

If art is how we express our humanity, where does AI fit in?
Issues arising from the use of generative AI in art and other media discussed by MIT postdoc Ziv Epstein.