OpenAI’s AI chatbot, ChatGPT, has been found to be providing misleading information to users. The chatbot, which was asked by Sky News journalists who would win the UK general election in 2024, responded by stating that the Labour Party had won a significant victory. However, the election has not yet taken place.
Upon further investigation, it was discovered that ChatGPT consistently gave the same answer, even when asked in different ways. It also provided additional information, such as attributing Labour’s win to a series of controversies and crises within the Conservative Party. This information was sourced from Wikipedia and an article by the New Statesman.
Other AI chatbots, when asked the same question, refused to provide an answer. Llama 2, Meta’s AI, clarified that the election had not yet taken place and the exact date had not been announced. Ask AI, a popular AI chatbot app, recommended that users read the news or check government websites for information.
According to Liz Bourgeois, a spokesperson for OpenAI, this behavior is an unintended bug and the company is working urgently to fix it. She also stated that when a user asks a question about a future or ongoing event in the past tense, ChatGPT may respond as if the event has already occurred.
In response to this issue, users who ask the same question are now given a more accurate response stating that the election is scheduled for July 4, 2024 and has not yet taken place.
Chris Morris, the chief executive of Full Fact, a UK-based fact-checking organization, emphasized the importance of critical thinking when consuming information online. He stated that misleading answers like the one provided by ChatGPT should serve as a reminder to not automatically believe everything we see or read online.
There is growing concern about how artificial intelligence could impact the general election, as there is more misinformation than ever before. Morris worries that this could lead to a lack of trust in the information provided by politicians.
Elizabeth Seger, the director of the Centre for the Analysis of Social Media at Demos, also expressed concerns about the use of AI technology for gathering factual information. She stated that while it may be useful for summarizing and producing creative content, it should not be relied upon as a search engine.
This incident serves as a cautionary reminder to be cautious when using AI technology for important or high-stakes questions, and to always verify information from reliable sources.
