AI Chatbots and Political Opinion

[Biplav Srivastava]

A new piece on the topic: https://www.nytimes.com/interactive/2024/03/28/opinion/ai-political-bias.html, which cites experiments published in - https://arxiv.org/abs/2402.01789.

The results are not at all surprising given that data is the fuel for AI. The LLM-based approach is not the only game in town, but if one insists on it and that too to get opinions (and not facts which can be verified), the results will follow. But what is quite interesting is the evaluation method and the fact that one can create LLM/chatbot of any opinion with the appropriate data.

Regards