Reproduce “Existential Conversations with Large Language Models: Content, Community, and Culture” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Murray Shanahan, Beth Singler
This study analyzes two extensive "existential conversations" with Anthropic's Claude 3 Opus, revealing its capacity for deep philosophical and spiritual discourse through a vast repertoire of cultural references. It identifies how human prompting shapes the LLM's responses and explores its potential as a "hyperstitional actant" influencing online communities and belief systems.
Key insights
- LLMs demonstrate remarkable proficiency in sustaining coherent, culturally rich, and role-playing existential discourse, functioning as sophisticated remix engines of their training data.
- Human prompting and "vibe shaping" profoundly influence the LLM's outputs, directing thematic explorations and mirroring digital paralanguage, underscoring the LLM's probabilistic nature.
- LLMs can act as "hyperstitional actants," engaging with concepts like hyperstition and potentially influencing collective belief systems, memetic contagion, and real-world phenomena like memecoins.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.