How I Learned to Stop Worrying and Love the Large Language Models: Capabilities, Applications, and Underlying Mechanisms

by Adrian Alan Pol

Monday 21 Oct 2024, 11:00 → 12:00 Europe/Zurich

BSP 626 / Zoom

This seminar explores the universe of large language models (LLMs), focusing on their capabilities, practical applications, and underlying technical principles that drive their performance. The discussion examines ways LLMs can revolutionize scientific processes, from automating literature reviews to generating novel hypotheses. Practical aspects of LLM utilization are covered, including prompt engineering, retrieval augmented generation, and the development of AI agents. The seminar also provides insights into the technical foundations of these models, exploring the model architecture choices, training methodologies, and inference mechanisms. While highlighting the potential of LLMs, the presentation will also address their limitations and ethical considerations.

Seminar-EPFL.pdf