Unlocking LLM Potential: From Federated Learning to Introspection

We have scheduled two talks on different parts of LLM. Please register at on AICamp (our event partner) website to get zoom link.

Agenda
12 pm -- 12:05 pm member join online
12:05 pm -- 12: 50 pm -- Talk 1 (Holger Roth, Nvidia)
12:50 pm -- 1:35 pm -- Talk 2 ( Jason Toy)
1:35 pm -- 1:40 pm closing

Talk 1: Empowering Federated Learning for Massive Models with NVIDIA FLARE
In the ever-evolving landscape of artificial intelligence (AI) and large language models (LLMs), handling and leveraging data effectively has become a critical challenge. Most state-of-the-art machine learning algorithms are data-centric. However, as the lifeblood of model performance, necessary data cannot always be centralized due to various factors such as privacy, regulation, geopolitics, copyright issues, and the sheer effort required to move vast datasets. In this talk, we explore how federated learning enabled by NVIDIA FLARE can address these challenges with easy and scalable integration capabilities, enabling parameter-efficient and full supervised fine-tuning of LLMs for natural language processing and biopharmaceutical applications to enhance their accuracy and robustness (For details, see our paper).

Speaker: Holger Roth
Holger Roth, a Principal Federated Learning Scientist at NVIDIA, specializes in developing distributed and collaborative software and models for various industries using federated learning and analytics. He has been exploring the topic both from theoretical and practical standpoints. During the COVID-19 pandemic, he led the experimentation of a federated learning study involving twenty hospitals around the globe to train more generalizable models for predicting clinical outcomes in symptomatic patients. His other research interests include computer-assisted annotation, active learning, and natural language processing. He is an Associate Editor for IEEE Transactions of Medical Imaging and holds a Ph.D. from University College London, UK. In 2018, he was awarded the MICCAI Young Scientist Publication Impact Award.

Talk 2: Metacognition is all you need? - Using Introspection in Generative Agents to Improve Goal-directed Behavior

A review of the paper "Metacognition is all you need? ": Recent advances in Large Language Models (LLMs) have shown impressive capabilities in various applications, yet LLMs face challenges such as limited context windows and difficulties in generalization. In this paper, we introduce a metacognition module for generative agents, enabling them to observe their own thought processes and actions. This metacognitive approach, designed to emulate System 1 and System 2 cognitive processes, allows agents to significantly enhance their performance by modifying their strategy. We tested the metacognition module on a variety of scenarios, including a situation where generative agents must survive a zombie apocalypse, and observe that our system outperform others, while agents adapt and improve their strategies to complete tasks over time.

Speaker : Jason Toy

Jason Toy is part of an AI research group building the best open source agent simulation framework at [https://replicantlife.com](https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Freplicantlife.com%2F&data=05%7C02%7Cchesterc%40nvidia.com%7C9e0bbc6c685449ca1fef08dc317979fa%7C43083d15727340c1b7db39efd9ccc17a%7C0%7C0%7C638439645014099040%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C0%7C%7C%7C&sdata=zZplSh3ypTV4Gs4PnTmFEc0P8BGI98wnzNU9VqXxjQ0%3D&reserved=0) . He has built and ran multiple engineering teams.