Thomas Lumley • 8/15/2025

Interviewing your laptop

The article critically examines the idea of using large language models (LLMs) like ChatGPT to replace traditional human surveys for opinion polling. It argues that LLMs face two core problems: accurately representing diverse, real-world demographics and correctly weighting those demographics to reflect a population. Using a science-fiction analogy, it illustrates why AI cannot solve the fundamental challenges of polling, especially for groups that are not 'Very Online'.

0 comments

#llm #large language models #ai ethics