船井業績賞記念講演
船井業績賞記念講演
Responsible Adaptation of Large Generative AI Models for Domain Specific Learning
Ling Liu
Professor, School of Computer Science, Georgia Institute of Technology
Professor, School of Computer Science, Georgia Institute of Technology
ABSTRACT/SUMMARY：The human-like generative ability of LLMs has ushered in a new era of foundational models and generative AI (genAI), unlocking new possibilities and driving cross-domain innovations. However, the transformative potential of these genAI models is hindered by significant accessibility challenges: (i) Powered by over-parameterization, LLMs are requiring hundreds of GBs of GPU memory for learning and inference, hence facing deployment challenges on heterogeneous platforms and on learning for downstream tasks with proprietary data, making equitable accessibility of genAI for all a grand challenge. (ii) Large genAI models trained on massive public domain data may introduce problematic hallucinations, which can lead to misinformation and biased outcomes in mission-critical applications, making responsible adaptation of genAI models another grand challenge.
This keynote will present a responsible and resource efficient framework for adapting genAI to domain-specific learning, aiming to tackle the above mentioned two accessibility challenges. I will first review the pros and cons of existing augmentation generation techniques for mitigating hallucination-induced misinformation and inaccuracies and introduce the multi-genAI-agent collaboration framework for responsible deployment of genAI in mission critical applications. Next, I will describe scalable and resource-efficient federated finetuning framework and optimizations for learning downstream tasks on proprietary or privacy-sensitive datasets with a population of heterogeneous clients.
This keynote will present a responsible and resource efficient framework for adapting genAI to domain-specific learning, aiming to tackle the above mentioned two accessibility challenges. I will first review the pros and cons of existing augmentation generation techniques for mitigating hallucination-induced misinformation and inaccuracies and introduce the multi-genAI-agent collaboration framework for responsible deployment of genAI in mission critical applications. Next, I will describe scalable and resource-efficient federated finetuning framework and optimizations for learning downstream tasks on proprietary or privacy-sensitive datasets with a population of heterogeneous clients.
Biography：Ling Liu is a Professor in the School of Computer Science at Georgia Institute of Technology. She directs the research programs in the Distributed Data Intensive Systems Lab (DiSL), examining various aspects of Internet-scale big data powered artificial intelligence (AI) systems, algorithms and analytics, including performance, reliability, privacy, security and trust. Prof. Liu is an elected IEEE Fellow, a recipient of IEEE Computer Society Technical Achievement Award (2012), and a recipient of the best paper award from numerous top venues, including IEEE ICDCS, WWW, ACM/IEEE CCGrid, IEEE Cloud, IEEE ICWS. Prof. Liu served on editorial board of over a dozen international journals, including the editor in chief of IEEE Transactions on Service Computing (2013-2016), and the editor in chief of ACM Transactions on Internet Computing (since 2019). Prof. Liu is a frequent keynote speaker in top-tier venues in Big Data, AI and ML systems and applications, Cloud Computing, Privacy, Security and Trust. Her current research is primarily supported by USA National Science Foundation under CISE programs, CISCO and IBM.