During the recent United States presidential election, AI chatbot services were closely scrutinized for their ability to provide accurate information related to the election or any biases present. This led many developers to opt for different approaches.
OpenAI revealed that its ChatGPT chatbot employed a question-filtering method regarding the US election, providing links to external websites for further information. These links included explanations of voting processes or latest election results from reputable news sources like Reuters and AP. OpenAI stated that they received over 2 million questions about election results on the day of the election and rejected over 250,000 requests to generate fake politician images.
Perplexity, an AI search service, took a different approach by creating a comprehensive website with election data and using AI chatbots to answer questions based on this data, primarily referencing information from AP. They reported over 4 million website visits on election day and no issues with providing inaccurate responses.
On the other hand, the chatbots Claude from Anthropic and Gemini from Google took a more direct approach by outright refusing to answer questions about the US election.
TLDR: AI chatbot services were under scrutiny during the US presidential election, with OpenAI, Perplexity, Anthropic, and Google’s chatbots employing different strategies in handling election-related queries.
Leave a Comment