Microsoft has acknowledged the issue and stated that it has taken steps to improve its conversational AI platforms, particularly in anticipation of the 2024 US elections. The company is focusing on authoritative sources of information for Copilot and encourages users to exercise their best judgment when viewing results. The potential for AI to mislead voters is a growing concern, and Microsoft has expressed a desire to work with political parties and candidates to limit deepfakes and prevent election misinformation.
Key takeaways:
- Microsoft's chatbot, Bing/Copilot, provided false and misleading information about European elections, according to a report by AlgorithmWatch.
- One-third of the chatbot's answers to election-related questions had factual errors, including fake controversies, wrong election dates, incorrect polling numbers, and candidates who weren't running in the elections.
- Microsoft has implemented guardrails on the chatbot to prevent it from providing dangerous, false, or offensive answers, but the chatbot still evaded questioning 39% of the time and only 30% of the answers were judged as factually correct.
- Microsoft has acknowledged the issue and is taking steps to improve its conversational AI platforms, especially ahead of the 2024 elections in the United States, focusing on authoritative sources of information for Copilot.