May 27, 2024 – Google’s newly launched “AI Overviews” feature in its search engine has been under fire recently due to its frequent provision of grossly inaccurate search results. One notable example saw the feature suggesting the use of glue to prevent cheese from sliding off pizzas, sparking widespread criticism and concern.
Earlier this week, The Verge reported that Google CEO Sundar Pichai acknowledged in an interview that the “hallucinations” produced by the AI Overviews feature are an “inherent flaw” of large language models (LLMs), which form the core technology behind the feature. Pichai stated that this issue remains an unsolved problem, indicating that despite the efforts of Google engineers to address the strange and critically erroneous answers generated by AI Overviews, such problems are likely to persist.
However, Pichai appeared to downplay the severity of these errors, arguing that occasional mistakes do not undermine the usefulness of the feature. He emphasized the progress made in terms of factual accuracy compared to the previous year and expressed optimism about the overall utility of AI Overviews.
Nonetheless, the errors made by the feature have caused a significant stir online, with many users sharing examples of incorrect information generated by AI Overviews. This has further damaged the credibility of Google’s search engine, which has already faced criticism for providing low-quality results in the past.
It is worth noting that AI advisor and SEO expert Branie Mueller commented on social media, expressing concern that the accuracy of AI is not always superior to traditional methods. She criticized Google for taking a risky gamble in the search field to surpass competitors like Perplexity and OpenAI, suggesting that they could have used AI for more valuable applications.