After Google’s AI generated numerous bizarre answers for users–what in AI terms is called “hallucinations– the tech giant rushed to quickly fix this string of incorrect and possibly dangerous results.
The tool, called AI Overviews, was introduced in the U.S. this month and is now facing significant backlash for spreading misinformation, including one that outrageously advised users to eat rocks for nutrition.
As part of a satirical post, The Onion posed a query to the AI tool asking “how many rocks should a child eat?” To which it responded by recommending “eating at least one small rock per day,” falsely citing UC Berkeley geologists.
Aside from this instance of false information that has gained traction online, the chatbot has generated many other strange responses such as telling users to add glue to pizza sauce and saying that tobacco has health benefits for kids.
The chatbot also responded to a question, “How many Muslim presidents has the U.S. had?” by falsely stating that it has “had one Muslim president, Barack Hussein Obama.”
In response to the incorrect answers produced, Google said it is “taking swift action where appropriate under our content policies, and using these examples to develop broader improvements to our systems, some which have already started to roll out.”
“The vast majority of AI Overviews provide high quality information, with links to dig deeper on the web,” the Google spokesperson said in a statement. “Many of the examples we’ve seen have been uncommon queries, and we’ve also seen examples that were doctored or that we couldn’t reproduce.”
Despite Google’s argument that these are isolated instances, many researchers of AI have warned that a major risk with rapidly growing technology is the spread of misinformation from false claims known as “hallucinations.”
“Google Search is Google’s flagship product and cash cow,” said Gergely Orosz, author of the Pragmatic Engineer Newsletter. “It’s the one property Google needs to keep relevant/trustworthy/useful. And yet, examples on how AI Overviews are turning Google Search into garbage are all over my timeline.”
Tech journalist Brian Merchant posted a screenshot in which Google’s own AI search results claimed that the company has violated antitrust law, citing the Justice Department’s pending lawsuits against the company. Google’s other AI chatbot, Bard, made the same mistaken claim last November.
