 "Google to improve AI Overviews, impose limits after odd, incorrect answers")
Google is placing limits on its AI Overviews search feature after it produced several bizarre and inaccurate results, the company said on May 30.
AI Overviews, introduced at Google I/O 2024 on May 14 after nearly a year of experimentation, offers users quick AI-generated summaries of topics with links for deeper exploration, appearing at the top of search results, according to Moneycontrol.
Initially made available to ‘hundreds of millions’ of users in the US, the tech giant plans to extend AI Overviews to over a billion people by the end of the year. This launch is part of Google’s effort to revamp its search product for the generative AI era, amid renewed competition from companies like Microsoft, OpenAI, and new entrants like Perplexity.
Google’s AI accused of ‘hallucination’
However, in the past week, the feature reportedly produced some unusual and incorrect advice, such as suggesting users eat rocks or use glue on pizza to help cheese stick, screenshots of which circulated widely online.
In a blog post on May 30, Google’s head of search, Liz Reid, said the company has now implemented detection mechanisms for ‘nonsensical queries’ and has limited the inclusion of satire and humour content in AI Overviews.
Also Read
Additionally, Google has updated its systems to reduce the use of user-generated content that might offer misleading advice and added restrictions for queries where AI Overviews were not particularly helpful. The company is also addressing the ‘small number of AI Overviews that violate content policies’, which includes potentially harmful or obscene information.
Reid noted that policy violations were found in “less than one in every 7 million unique queries” featuring AI Overviews. She argued that AI Overviews generally do not ‘hallucinate’ or invent information; errors usually stem from misinterpreting queries, nuances of language, or insufficient information.
Google’s AI Overview enhancements
Reid compared AI Overviews to ‘featured snippets’, a long-standing search feature that uses AI to highlight key information with links to web content, claiming their accuracy rates are comparable. She emphasised that the feature underwent extensive testing before launch, including rigorous red-teaming, evaluations with typical user queries, and tests on a portion of search traffic.
Furthermore, Reid stated that Google will not show AI Overviews for hard news topics where timeliness and accuracy are crucial, and for certain health topics, the company has added additional refinements to better its quality protections.
These updates come as Google aims to monetise its AI features by testing search and shopping ads within AI Overviews.