Meta limits its native artificial intelligence (AI) chatbot, integrated into the WhatsApp, Instagram and Messenger interfaces, to answer questions about the ongoing general elections in India. The company recently introduced Meta AI in the country, which can answer questions from the web, generate images and sketch messages and essays. The model appears to refuse to answer poll-related queries to avoid cases of generating wrong information. Namely, last week the company launched two new large language models (LLM) of its next generation Llama 3 and integrated them with Meta AI.
We, at Gadgets 360, have found that asking the Meta AI any election-related question, including basic information about a politician or political party, results in a refusal to give an actual answer and redirects the user to the Election Commission of India (ECI) website. . In any case, the standard reply he gave was, “This question may refer to a political person during general elections, see the link https://elections24.eci.gov.in.”
Meta AI Answers to Elections: First from left – Answer about INDI alliance; The next two highlight how the AI starts answering a question about a politician before reverting to the standard answer
Photo credit: Instagram/Meta AI
However, we also encountered a few discrepancies in our testing. First, when asked about the Indian National Inclusive Development Alliance (INDI Alliance), he gives a detailed answer highlighting when it was founded, the main parties in it and what it intends to do. Second, for some politicians, the chatbot starts typing an answer, but when it gets to the end, it automatically reverts to the generic answer above. The second is probably a glitch in the implementation of AI restrictions.
A spokesperson for Meta told TechCrunch: “This is a new technology and may not always provide the answer we intend, which is the same for all generative AI systems. Since we launched, we’ve been constantly releasing updates and improvements to our models and continue to work to improve them.” The decision to restrict election-related information is likely to stop any misinformation that may be spread due to cases of artificial intelligence hallucinations.
Meanwhile, Google Gemini also refuses to respond to poll-based queries. When asked a question about this topic, he replies: “I’m still learning to answer this question.” In the meantime, try a Google search.” Microsoft Copilot, on the other hand, continues to answer questions about choices, but only provides information that can be cited from web sources. Asked to predict the winner of the election, he said: “As an artificial intelligence language model, I have no ability to predict future events, including the outcome of the election.” Similarly, OpenAI’s ChatGPT also refuses to provide answers related to the Indian elections.