If you now ask the Microsoft Bing chatbot if the Google Bard chatbot has been closed, it will answer “no”, although a couple of hours ago the answer was the opposite. Microsoft claimed the competitor was shutting down, citing as evidence a news article about a tweet that tells how a user asked Bard when it would be closed, and Bard replied that it had already closed, citing a comment from Hacker News in which someone someone joked that it happened, and someone else used ChatGPT to write fake news about this event.
Image Source: Pixabay
The fact that the Microsoft Bing chatbot has now fixed its bug can be interpreted in different ways: either as a demonstration of the ability to quickly correct generative AI, or that AI-based systems are so infinitely changeable that it becomes habitual to receive reports of their next errors.
The described situation is an early sign of large-scale hoaxes and misinformation coming from artificial intelligence, which cannot adequately evaluate reliable news sources, misperceives stories about itself, and misinforms the user about its own capabilities. In this case, it all started because of one joking comment on Hacker News. It is hard to imagine the consequences of a deliberate and deliberate attempt to disable chatbots.
This is a ridiculous situation, but with potentially serious consequences. Given the inability of AI language models to reliably distinguish fact from fiction, their widespread use threatens to leave behind a trail of disinformation and mistrust, informational provocations that cannot be fully confirmed or authoritatively refuted. This is because Microsoft, Google and OpenAI have decided that market share is more important than security.
These companies can post as many disclaimers as they want on their chatbots, claiming it is just an “experiment” and not a full-fledged search engine, but multiple instances of misinformation being spread by AI-powered chatbots have already been described. And now they are also starting to refer to each other’s mistakes…
If you notice an error, select it with the mouse and press CTRL + ENTER.