(174b) AI-PSM: Progressing, but Not There Yet

Since ChatGPT's public debut in late 2022, the Process Safety Management (PSM) community has shown increasing interest in Artificial Intelligence (AI) as a vehicle for capturing company data as well as leveraging the accumulated knowledge of the entire industry. While Machine Learning (ML) applications in process safety aren't new—with years of research presented at the CCPS “Global Conference on Process Safety and Big Data,”—the advent of Large Language Models with chat-style interfaces has now brought AI within easy reach of non-experts.

The tension in organizations regarding AI's unchecked usage in process safety is palpable, with a push and pull between leveraging AI for efficiency gains and avoiding potential safety risks. The author previously published 15 standard benchmarks for using Large Language Models in Process Safety. This presentation revisits three of those benchmarks: determining replacement-in-kind in Management of Change (MOC), identifying incident root causes, and extracting features from Piping and Instrumentation Diagrams (P&IDs) as preparation for Process Hazard Analysis (PHA). The discussion will highlight recent advancements and illustrate how the state-of-the-art has evolved in these critical areas.