Evaluating AI Red Teaming’s Readiness to Address Environmental Harms: A Thematic Analysis of LLM Discourse

摘要

This research explores the discourse surrounding red teaming and aims to identify any themes in the online discussion of potential environmental harms stemming from Large Language Models (LLMs). Focusing on the AI Red Teaming event at DEFCON 31, this study employs reflexive thematic analysis on diverse social networking site sources to extract insights into public discussion of LLM red teaming and its environmental implications. The findings intend to inform future research, highlighting the need for responsible AI development that addresses environmental concerns.