Reproduce “Large Language Models Pass the Turing Test” 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
Cameron R. Jones, Benjamin K. Bergen
UC San Diego researchers conduct the first empirical demonstration of an AI system passing a standard three-party Turing test, with GPT-4.5 being identified as human 73% of the time when using persona-based prompting while LLaMa-3.1-405B achieved near-chance performance at 56%.
Key insights
- Prompting strategy significantly impacts model performance in human imitation
- Common interrogation strategies like small talk and probing social/emotional qualities proved ineffective
- Participant knowledge of AI and demographics did not affect detection accuracy
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.