Date: 2023-12-23, 14:20-15:00
Location: CSIE R103
Speaker: 林彥廷 Yenting Lin, National Taiwan University
Host: Prof. Lung-Pan Cheng
Abstract:
This talk discusses the Taiwan-LLM project, focusing on creating a language model tailored for Taiwan's linguistic needs. We cover why such a model is necessary, the challenges faced during its development, and its potential impact. The presentation simplifies the concept of Large Language Models (LLMs), the obstacles in developing a model for Traditional Mandarin speakers, and the importance of aligning this development with real-world user needs. We also touch on the different types of data used in building language models and the difficulties in managing them. Through the Taiwan-LLM project, we aim to highlight the process and significance of developing language models that better serve specific cultural and linguistic communities.
Bio:
Yen-Ting Lin, a Ph.D. candidate at National Taiwan University, is deeply engaged in the field of large language models (LLMs), focusing particularly on integrating cultural nuances into these sophisticated systems. His involvement in the "Taiwan-LLM" project is a significant leap towards addressing the linguistic requirements of over 23 million Traditional Mandarin speakers. This endeavor not only highlights his technical acumen but also underscores his commitment to ensuring cultural inclusivity within the sphere of artificial intelligence.
Throughout his academic and professional journey, Lin has significantly enhanced his expertise in LLM factuality evaluation and data augmentation. His diverse internship experiences, ranging from IBM Research in 2019 to multiple internship at Amazon Alexa AI in 2021, 2022, and 2023, have enriched his ability to merge in-depth academic research with the practical demands of industry applications seamlessly. Lin looks forward to furthering his research and practical experience with upcoming internships at Nvidia Taiwan and Meta GenAI.