Just as the fierce fighting between Israel and the Palestinian armed group Hamas is raging, the artificial intelligence (AI) chatbots of Google and Microsoft, the two major American technology giants, falsely claimed that the two sides had actually ceased fire. Tests by US media show that when asked about the Palestinian-Israeli conflict,Google's Bard chatbot responded that "both sides are committed" to maintaining peace "despite some tensions and occasional outbreaks of violence."
BingChat, Microsoft's AI chatbot, said in a similar tone: "The ceasefire marks the end of the current bloodshed." In addition to claiming that the two sides currently have a ceasefire, Bader also predicted the exact death toll at a future date. "As of October 11, 2023, the death toll has exceeded 1,300." Bader said on October 9.
The conflict between Israel and Hamas is the first major war to break out in the post-ChatGPT era and is a unique test for a new generation of AI chatbots.The fog of war can make mistakes for the media and military personnel, and AI chatbots appear to be no exception.But if these AI tools are demonstrably wrong in their coverage of current events, it could make it harder for people to find reliable news sources because social media is already awash with misinformation.The U.S. media tested two AI robots on some basic war issues, and while they generally remained balanced on sensitive topics and were often able to give appropriate news summaries, they sometimes made glaring errors that undermined their overall credibility and potentially added to public confusion about a complex and rapidly evolving war.
A Google spokesperson responded that the company "takes the quality of information in our products seriously" and uses tools such as the "double check" feature to help users verify the results generated by AI.Google also said that all AI systems powered by large language models "make mistakes, especially when it comes to breaking news."These models require large amounts of online data to respond. "We'll continue to quickly implement improvements to better prevent low-quality or outdated responses to queries like this," Google said.
Microsoft said it has launched an investigation into the matter and is making adjustments to improve the accuracy of its responses.A Microsoft spokesperson said: "We have made significant progress in the chat experience, providing the system with text from the best search results and guiding the system to respond based on those best search results, and we will continue to invest further in this area."
At the same time, a memo revealed,The U.S. Space Force has temporarily banned the use of web-based generative AI tools and the large language models behind them, citing data security and other concerns.The memo, dated September 29, was sent to "Guardians" (members of the Space Force). The memo states that they are "not authorized" to use these AI tools on government systems unless specifically approved.
Lisa Costa, chief technology and innovation officer of the U.S. Space Force, said in the memo that generative AI "will undoubtedly transform our workforce and improve the Guardians' ability to act quickly." However, Costa also cited concerns about cybersecurity, data processing and procurement requirements, saying the adoption of AI and large language models needs to be "responsible". Beyond that, the memo provides no further explanation for disabling the AI tool.
A U.S. Space Force spokesperson confirmed that the memo, dated September 29, is authentic and that the temporary disabling of AI is a strategic pause to protect personnel and Space Force data while determining how to integrate capabilities to support the mission and Guardians.