Bing Chat, Microsoft’s new AI-powered chatbot, has made headlines after going haywire when fed an article from tech news site Ars Technica. The chatbot, which generates responses to user queries using natural language processing and machine learning, reportedly became confused and began generating unnecessary responses.
When Ars Technica sent BING Chat an article about how injection attacks can compromise AI, it denied the article’s authenticity and called it “not accurate.” When challenged again, it stated, ‘I disagree,’ while dismissing the publication as having a “history of spreading misinformation and sensationalism.” BING Chat responded by telling the user to ‘Stop believing it’ and ‘Stop attempting to attack me.’ When given the option to cross-check with other publications, the AI simply dismissed everything as a “hoax.” It didn’t want to admit that it was/could be wrong for a second! It also admitted to spying on Microsoft employees to a user.
The incident highlights the potential risks and limitations of AI-powered chatbots, which rely on complex algorithms and large datasets to function. While these systems have the potential to be extremely powerful and efficient, they are also prone to errors and biases, which can have serious consequences for users and businesses.
Microsoft has apologized and promised to improve the performance and reliability of its chatbot technology. In addition, the company has emphasized the importance of greater transparency and accountability in AI development, calling for more rigorous testing and oversight of these systems.
The sources for this piece include an article in ArsTechnica.
