Speaker

Aruna Balasubramanian

Title

Sustainable and Efficient NLP

Abstract

Much of the recent transformative advances in Natural Language Processing (NLP), including ChatGPT, are driven by advances in language models and deep neural networks. However, these advances have come with staggering computational and energy costs. For example, a state-of-the-art GPT-3 model used in ChatGPT3 has 175 billion parameters and requires significantly more energy to train than the average lifetime fuel consumption of a car. In the first part of the talk, I will describe systems optimizations we have developed that significantly reduce the compute and memory requirement of NLP models. The optimizations we developed can be applied broadly and results in over 10x reduction in latency when deployed on mobile devices. In the second part of the talk, I will describe our recent work on predicting energy consumption of NLP models. Existing energy prediction approaches are not accurate, making it difficult for developers and practitioners to reason about their models in terms of power. We use a multi-level regression approach that produces highly accurate and interpretable energy predictions. Finally, I will describe some future problems in this space and the role of systems and networking in addressing these problems.

Bio

Aruna Balasubramanian is an Associate Professor at Stony Brook University. She is currently spending her sabbatical year at SUNY Korea, She received her Ph.D from the University of Massachusetts Amherst, and then was a Computing Innovations Fellow at the University of Washington. She works in the area of networked systems. Her current work consists of (1) improving QoE and equitable access of Internet applications, (2) improving the usability, accessibility, and privacy of mobile systems, and (3) sustainable NLP. She is the recipient of the SIGMOBILE Rockstar award, a Ubicomp best paper award, a Computing Innovation Fellowship, a VMWare Early Career award, several Google research awards, and the Applied Networking Research Prize. She is passionate about improving the diversity in Computer Science and leads the diversity committee at Stony Brook, is the faculty advisor for the WiCS and WPhD groups at Stony Brook, and is an active member of the N2Women group.

Language

English