Inaccurate AI responses spark outrage
Google’s new AI search feature is under scrutiny for delivering false and bizarre answers, including claims that Barack Obama is Muslim and astronauts played with cats on the Moon. Launched in the U.S. last week and available in the UK since last month, the AI Overview tool aims to simplify information searches. However, its erratic behavior has prompted widespread criticism.
Bizarre and incorrect responses
Among the AI’s strange outputs, it advised users to mix non-toxic glue with cheese for pizza and claimed geologists recommend eating one rock per day. It also inaccurately stated that only 17 of the 42 U.S. presidents were white. These examples, sourced from Reddit and satirical sites like The Onion, have flooded social media, raising concerns about the feature’s reliability.
Google’s response and mitigation efforts
Google asserts that these are isolated incidents and do not reflect the tool’s overall performance. A spokesperson emphasized that most AI overviews provide high-quality information and that the company conducted extensive testing before launch. The tech giant is addressing policy violations and using these examples to refine its systems.
This incident is not unique to Google. Other AI products have also faced issues, such as ChatGPT fabricating a sexual harassment scandal and attributing it to a real law professor. More recently, OpenAI faced backlash from Scarlett Johansson for using a voice similar to hers without consent for its chatbot.
Google’s AI search troubles highlight ongoing challenges in ensuring the accuracy and reliability of AI-generated content.