AI and computing challenges: harnessing Large Language Models at Thomson Reuters

by Adrian Alan Pol (Thomson Reuters)

Monday 17 Jun 2024, 11:00 → 12:00 Europe/Zurich

40/S2-C01 - Salle Curie (CERN)

This talk will explore the transformative impact of large language models (LLMs) within the Thomson Reuters corporation. We will discuss automating content generation to empower legal, corporate, and tax professionals. We will examine the dual-edged nature of this technology, addressing the challenge of generating convincing yet inaccurate information, dubbed hallucinations, which necessitates rigorous validation mechanisms to ensure accuracy and reliability. Furthermore, we will analyze the substantial costs of running LLMs and explore strategies for optimizing cost and efficiency, considering perspectives from both Thomson Reuters and foundational model providers. The talk will also reflect on my journey from prototyping models for the CMS trigger system and collaborating with the industry through KT partnerships to working in a multinational technology corporation.

To be kept informed of KT Seminars please sign up at: http://cern.ch/go/F9cX

Stay up to date with CERN innovation partnerships and knowledge-transfer opportunities: http://kt.cern/newsletter

Follow us on LinkedIn: https://www.linkedin.com/showcase/cern-innovation-partnerships/

Recording