AI

Google's AI Overviews is being laughed at by the CEO of search rival Perplexity after the glue pizza fiasco

Selfie of a man, Aravind Srinivas, looking directly at the camera. Smiling with a head of black hair, wearing a blue Patagonia jacket.
Aravind Srinivas, cofounder and CEO of Perplexity AI. Aravind Srinivas
Read in app

Google's AI Overviews are wilding out.

Many users — including the CEO of its AI search rival Perplexity — have shared screenshots of some flawed responses it has generated.

Aravind Srinivas posted a laughing emoji in response to a post showing the suggestion to put glue on pizza. He also wrote "lel" on X and shared a screenshot of an AI Overviews error that says Batman is a cop. 

Srinivas also had a pop at Google's flub with Gemini's image-generating feature back in March.

Since Google launched its experimental tool at its I/O conference last week, people have been sharing what the AI search summary feature is getting wrong.

Recent examples of AI Overviews going haywire that were posted on X include responses claiming Barack Obama was a Muslim president, that Africa has no countries beginning with the letter K, and that people should eat "at least one small rock per day."

Google has said before that the incidents are "extremely rare queries and aren't representative of most people's experiences," but users continue to share their experiences with the inaccuracies it generates.

Some think Google may delay a wider rollout if it continues to produce such responses.

They threaten to dent Google's reputation, although it continues to dominate search, with an 86% market share in the US as of April, according to StatCounter. Competitors such as Microsoft's Bing are far behind at about 8%.

The inaccuracies that Google's AI search feature generates could also have harmful consequences, Melanie Mitchell, a professor at Santa Fe Institute with AI expertise, told Business Insider.

She said many people are concerned about AI becoming too intelligent, but that "the more immediate danger is AI that's too dumb being trusted to do a job it's not capable of."

She added: "This is what's happening with Google's AI Overview feature. I worry that because people generally trust Google searches, they will not question the (very untrustworthy) AI-generated overviews that Google now provides. An avalanche of misinformation, and disinformation, will be in store for all of us." 

Google and Perplexity AI didn't immediately respond to requests for comment from Business Insider, made outside normal working hours.

Correction: May 24, 2024 — A previous version of this story included an example that was not generated by Google AI Overviews.

Read next

Jyoti Mann
Jyoti Mann
Jyoti Mann is a former senior reporter for BI who covered tech companies, including OpenAI, Meta, Amazon, and Microsoft.Jyoti holds a master's degree in investigative journalism and was previously reporting on the companies desk at the Financial Times. She spent more than three years working as a hedge fund consultant carrying out investigative research. She has previously written for the Financial Times and The Times. Sample of stories: Amazon's full RTO is off to a bumpy start. Some staff complain of a lack of space and theft. And they're still on video chats.Internal Meta memo tells managers how performance-based job cuts will workStripe accidentally sent an image of a duck when notifying some employees they were getting laid offKlarna tells employees it will start drug testing workers in Sweden