Google criticized as AI Overview makes obvious errors, such as saying former President Obama is Muslim

News

In this article

Alphabet CEO Sundar Pichai speaks at the Asia-Pacific Economic Cooperation CEO Summit in San Francisco on Nov. 16, 2023.
David Paul Morris | Bloomberg | Getty Images

It’s been less than two weeks since Google debuted “AI Overviews” in Google Search, and public criticism has mounted after queries have returned nonsensical or inaccurate results within the AI feature — without any way to opt out.

AI Overviews show a quick summary of answers to search questions at the very top of Google Search: For example, if a user searches for the best way to clean leather boots, the results page may display an “AI Overview” at the top with a multi-step cleaning process, gleaned from information it synthesized from around the web.

But on social media, users have shared a wide range of screenshots showing the AI tool sharing controversial responses.

Google, Microsoft, OpenAI and other companies are at the helm of a generative AI arms race as companies in seemingly every industry race to add AI-powered chatbots and agents to avoid being left behind by competitors. The market is predicted to top $1 trillion in revenue within a decade.

Here are some examples of what went wrong with AI Overviews, according to screenshots shared by users.

When asked how many Muslim presidents the U.S. has had, AI Overviews responded, “The United States has had one Muslim president, Barack Hussein Obama.”

When a user searched for “cheese not sticking to pizza,” the feature suggested adding “about 1/8 cup of nontoxic glue to the sauce.” Social media users found an 11-year-old Reddit comment that seemed to be the source.

For the query “Is it OK to leave a dog in a hot car,” the tool at one point said, “Yes, it’s always safe to leave a dog in a hot car,” and went on to reference a fictional song by The Beatles about it being safe to leave dogs in hot cars.

Attribution can also be a problem for AI Overviews, especially when it comes to attributing inaccurate information to medical professionals or scientists.

For instance, when asked “How long can I stare at the sun for best health,” the tool said, “According to WebMD, scientists say that staring at the sun for 5-15 minutes, or up to 30 minutes if you have darker skin, is generally safe and provides the most health benefits.” When asked “How many rocks should I eat each day,” the tool said, “According to UC Berkeley geologists, people should eat at least one small rock a day,” going on to list the vitamins and digestive benefits.

The tool also can respond inaccurately to simple queries, such as making up a list of fruits that end with “um,” or saying the year 1919 was 20 years ago.

When asked whether or not Google Search violates antitrust law, AI Overviews said, “Yes, the U.S. Justice Department and 11 states are suing Google for antitrust violations.”

The day Google rolled out AI Overviews at its annual Google I/O event, the company said it also plans to introduce assistant-like planning capabilities directly within search. It explained that users will be able to search for something like, “‘Create a 3-day meal plan for a group that’s easy to prepare,'” and they’d get a starting point with a wide range of recipes from across the web.

Google did not immediately return a request for comment.

The news follows Google’s high-profile rollout of Gemini’s image-generation tool in February, and a pause that same month after comparable issues.

The tool allowed users to enter prompts to create an image, but almost immediately, users discovered historical inaccuracies and questionable responses, which circulated widely on social media.

For instance, when one user asked Gemini to show a German soldier in 1943, the tool depicted a racially diverse set of soldiers wearing German military uniforms of the era, according to screenshots on social media platform X.

When asked for a “historically accurate depiction of a medieval British king,” the model generated another racially diverse set of images, including one of a woman ruler, screenshots showed. Users reported similar outcomes when they asked for images of the U.S. founding fathers, an 18th-century king of France, a German couple in the 1800s and more. The model showed an image of Asian men in response to a query about Google’s own founders, users reported.

Google said in a statement at the time that it was working to fix Gemini’s image-generation issues, acknowledging that the tool was “missing the mark.” Soon after, the company announced it would immediately “pause the image generation of people” and “re-release an improved version soon.”

In February, Google DeepMind CEO Demis Hassabis said Google planned to relaunch its image-generation AI tool in the next “few weeks,” but it has not yet rolled out again.

The problems with Gemini’s image-generation outputs reignited a debate within the AI industry, with some groups calling Gemini too “woke,” or left-leaning, and others saying that the company didn’t sufficiently invest in the right forms of AI ethics. Google came under fire in 2020 and 2021 for ousting the co-leads of its AI ethics group after they published a research paper critical of certain risks of such AI models and then later reorganizing the group’s structure.

Last year, Pichai was criticized by some employees for the company’s botched and “rushed” rollout of Bard, which followed the viral spread of ChatGPT.

Articles You May Like

Microsoft Launches Windows Resiliency Initiative to Boost Security and System Integrity
Apple and Google could face a competition probe over their huge mobile ecosystems in the UK
Microsoft, Meta, and DOJ Disrupt Global Cybercrime and Fraudulent Networks
Indian Government to Offer Up to $5 Billion in Incentives for Electronics Production, Weaning Off China
iPhone 17 Air to Be Apple’s Thinnest-Ever Phone; A19 Chips to Be Built Using TSMC’s New Technology: Report