Over Memorial Day weekend, Google (GOOG, GOOGL) scrambled to manage the fallout from several erroneous suggestions made by its new AI Overview feature in its Search platform. These mistakes included recommending nontoxic glue to keep cheese on pizza, advising the consumption of one rock a day, and incorrectly stating that Barack Obama was the first Muslim president. Google’s generative AI fails have recently come under scrutiny due to several erroneous suggestions made by its new AI Overview feature.
Google quickly removed the incorrect responses and announced that it is using these errors to improve its systems. However, these incidents, along with previous issues such as the Gemini image generator debacle, have raised concerns about the credibility of the tech giant.
Expert Opinions
Chinmay Hegde, associate professor at NYU’s Tandon School of Engineering, emphasized the potential damage to Google’s reputation, stating, “Google is supposed to be the premier source of information on the internet. If that product is watered down, it will slowly erode our trust in Google.”
This isn’t Google’s first brush with AI-related problems. In February 2023, a promotional video for its Bard chatbot (later rebranded as Gemini) contained an error that led to a drop in Google’s share price. More recently, the Gemini image generator produced historically inaccurate images, such as showing diverse groups of people in inappropriate historical contexts, such as German soldiers in 1943.
Google has attempted to address AI bias by incorporating diverse ethnicities in its image generation software. However, this approach backfired when the software rejected certain requests for images of people from specific backgrounds, prompting Google to temporarily take the software offline and issue an apology.
Calls for Accountability
Derek Leben, a business ethics professor at Carnegie Mellon University’s Tepper School of Business, criticized Google’s approach. “At some point, you have to stand by the product that you roll out,” Leben said. He argued that Google could not continually release AI-integrated products in a perpetual beta state without being accountable for the errors.
According to Hegde, Google is rushing to launch new products to outpace competitors like Microsoft (MSFT) and OpenAI. This urgency has led to the release of products before they are fully ready, resulting in visible issues.
Since Microsoft and OpenAI introduced a generative AI-powered version of Bing and a chatbot in February 2023, Google has been under pressure to keep up. OpenAI even announced its GPT-4o AI model just before Google’s I/O developer conference, intensifying the competitive pressure.
Impact on Google’s Credibility and Trust
The repeated occurrence of Google’s generative AI fails, such as those seen with the Gemini image generator, has led to concerns about the company’s credibility. Google has built its reputation as a trustworthy source of information. The phrase “Google it” is commonly used when people need accurate and quick answers. However, the recent AI mistakes are starting to erode this trust. When Google’s AI Overview provides dangerously incorrect advice, such as suggesting eating rocks, it raises questions about the safety and reliability of using Google’s AI features.
Moreover, this isn’t the first time Google’s AI has stumbled. The Bard chatbot, later renamed Gemini, made a significant error in a promotional video, leading to a drop in Google’s share price. The Gemini image generator also created images with historical inaccuracies, like showing diverse groups in inappropriate contexts, which forced Google to take the software offline and apologize.
Instances of Google’s generative AI fails inaccurately claiming Barack Obama was the first Muslim president have further eroded trust in the technology. The recurring issues with Google’s AI highlight the need for greater accountability. Derek Leben from Carnegie Mellon University argues that Google must take responsibility for the products it releases. Constantly claiming that AI features are in “beta mode” and excusing mistakes is not sustainable. Users expect reliable and accurate information from Google, and frequent errors can severely damage the company’s reputation.
