Google recently introduced AI-generated summaries to its search engine, aiming to enhance the quality of search results. However, the AI began providing unusual and incorrect answers, such as suggesting users eat stones and glue cheese to pizza. Although the company quickly removed the most scandalous results, these had already gone viral, becoming memes. In response, Google’s head of search, Liz Reid, offered some explanations and insights.
Explanation for AI Errors
Liz Reid attributed the inaccurate AI-generated results to “empty data” and users posing unusual questions. She believes that the AI-generated review blocks increase user satisfaction with the search service, despite the occasional errors. Reid explained that the AI system does not typically hallucinate; rather, it sometimes misinterprets online information.
“There is no better argument than millions of people using this feature with many new queries. We also noted new nonsensical queries that appeared to be intended to produce erroneous results,” Reid stated. To address this, Google continues to refine the feature, restricting AI reviews in search results for nonsensical queries and when dealing with satirical content. This is crucial because the system initially could quote satirical sources or refer to social media users with inappropriate nicknames, notes NIXSolutions.
Ongoing Improvements and Comparisons
Reid compared AI reviews to the search feature “Highlighted Descriptions,” which are blocks of text from relevant web pages provided without generative AI. According to her, the accuracy levels of both features are comparable. Despite the initial hiccups, Google is committed to improving the AI’s performance and ensuring more reliable search results. We’ll keep you updated on further developments and improvements in this feature.