![Google puts limits on AI Overviews](/media/k2/items/src/Google-on-AI-Overviews.jpg)
(Image source from: REUTERS)
Google announced on May 30 that it would impose restrictions on its artificial intelligence search experience, AI Overview, after it gave consumers strange and incorrect results. After nearly a year of testing, Google announced an overview of its AI at its annual developer conference Google I/O 2024 on May 14. This feature provides consumers with a quick AI-generated overview of a topic and links for deeper exploration at the top of search results. AI Overviews is now available to “hundreds of millions” of users in the US and plans to expand to more than a billion users by the end of the year. The launch comes as the tech giant looks to reimagine its flagship search product for the age of generative AI in the face of new competition from rivals like Microsoft and OpenAI, as well as startups like Perplexity. But last week, the search function gave consumers strange and incorrect results, including telling them to eat stones and put glue on the pizza to better remove the cheese.
Google search chief Liz Reed said in a May 30 blog post that the company is currently building a mechanism to identify "nonsense searches" that shouldn't display AI summaries, citing publishing limitations. The tech giant also updated its systems to limit the use of user-generated content in answers that could give misleading advice and for questions where AI summaries proved less useful than Trigger's limits, he said. Google is also taking action against “a small number of AI digests that violate our content policies,” Reed said. This includes summaries that contain potentially harmful, offensive or infringing information. Reed said the AI summary found content policy violations in "fewer than 1 in 7 million unique requests" served. The developments also come as Google seeks to monetize its AI services by testing search and shopping ads in the AI dashboard.
In a blog post, Reed blamed these false results on an "information vacuum" or "information gap" where the quality of content on a topic is limited. “Until these screenshots went viral, almost no one asked Google this question, and a lot of content on the web didn’t take it seriously,” he says. Reid said there is humorous content about a particular rock sample that is "randomly reposted on a geology software provider's website so that someone searching for this question can get an overview of the information.""See Artificial. We honestly linked to one of the few websites that deals with this." In some cases, such as the pizza example, Reid says AI Overviews misinterprets the language of web pages and displays sarcastic content from forums: "In In some cases, it's a great source of real, first-hand information, but in some cases, says Reid, it can lead to advice that isn't very helpful, like using glue to put cheese on a pizza to stick.