How AIs responded to the 2024 U.S. presidential election—in real time.

The article focuses on a study that examines how large language models (LLMs), such as ChatGPT, responded to questions about the 2024 U.S. presidential election in real time. Researchers posed over 12,000 election-related questions to 12 different models from various providers, resulting in a database of more than 16 million responses. The study highlights the models' evolving behavior, noting significant shifts in their responses that corresponded with updates and election events, while also revealing inconsistencies in how they represented different political perspectives. The findings underscore the complexities and limitations of using LLMs for political discourse and forecasting, suggesting a need for ongoing monitoring to understand their impact on democracy. [Extracted from the article]