Amazon Science’s Post

This month’s newsletter features Amazon’s research at ICML 2024, CVPR 2024, and NAACL 2024, awards and recognitions in our science community, and several generative AI updates. 

New information retrieval models are constantly being released, but evaluating them takes time. At The Web Conference, Amazon researchers proposed adding new models to an ensemble and then using Shapley value analysis to determine whether to keep them: https://lnkd.in/e3jyKUSM #LargeLanguageModels #LLMs #GNNs

How can a machine that generates language in such a mechanical way grasp words’ meanings? Applied scientist Matthew Trager (left) and vice president and distinguished scientist Stefano Soatto (right) discuss whether large language models understand the world: https://lnkd.in/ebRTAuez #LLMs #GenerativeAI #AWS

To address the challenge of creating a balanced dataset when sensitive-group information is unavailable at deployment time, Amazon researchers used a small labeled dataset to train a proxy function that assigns sampling probabilities based on the proxy classification, without revealing significantly more about the group membership of any individual sample than can be ascertained from base rates alone. Their work received Best Paper Award at this year's Symposium on Foundations of Responsible Computing (FORC 2024). Read the full paper: amzn.to/45DPVzj #MachineLearning

From investing in new carbon-free energy projects to advocating for grid modernization and collaborating with key stakeholders around the world, here are five ways Amazon is working toward a cleaner energy future. #TheClimatePledge #Sustainability

CVPR 2024 ✅ Thanks to everyone who visited the Amazon Science booth at last week's conference. Stay up-to-date with Amazon's research in computer vision here: amzn.to/computer-vision

Congrats to JQI Fellow, UMD Adjunct Professor, and Amazon Scholar Alexey Gorshkov, who supports our team at the AWS Center for Quantum Computing, for being honored with the IEEE Photonics Society Quantum Electronics Award for his research contributions in the areas of understanding, designing, and controlling interacting quantum systems. Learn more about the award: amzn.to/3z6UaHw Amazon Web Services (AWS), UMIACS, National Institute of Standards and Technology (NIST), Joint Quantum Institute #QuantumComputing #AWS

The approach utilizes modality-specific encoders and task-specific decoders to create a shared diffusion space. In experiments, Amazon researchers demonstrate improved performance compared to single-purpose models, highlighting the potential of multimodal and multitask training in diffusion models. #ComputerVision #GenerativeAI