Google’s search AI has been in public beta for nearly a year now, but it seems like it’s still struggling to get things right. The Washington Post recently delved into Google’s Search Generative Experience (SGE) and found that, despite some improvements, the AI-powered update is still serving up incorrect and sometimes downright bizarre answers. While SGE has thankfully stopped suggesting absurd things like melting eggs or praising slavery, it’s now facing criticism for what experts term “hallucinating” – essentially, making things up. For instance, a search for a fictional Chinese restaurant in San Francisco called “Danny’s Dan Dan Noodles” yielded results about long lines and crazy wait times, complete with made-up citations about 4,000-person queues and a two-year waitlist.
But hallucinations aren’t the only issue plaguing SGE. An analysis by the SEO platform SE Ranking revealed that the AI-informed search beta heavily relies on Quora, a platform notorious for dubious information, as its most-cited source. This reliance on questionable sources has led to some glaring inaccuracies, such as the mishandling of financial data. In a comparison between regular Google search and SGE, when asked about Meta CEO Mark Zuckerberg’s net worth, the former provided a straightforward answer of $169.8 billion USD. However, SGE’s response was a convoluted breakdown, stating Zuckerberg’s net worth as “$46.24 per hour, or $96,169 per year,” escalating to “$230.6 million per day.” While Google has since updated some of its data sources and rectified the issue, the incident underscores the challenges inherent in relying on AI for accurate information.
Moreover, Google’s handling of SGE has raised concerns about user consent and transparency. Despite Google’s assertion that SGE is opt-in only, Search Engine Land revealed that the generative AI feature is being made default for users who haven’t explicitly chosen to use it. This move has sparked criticism over Google’s approach to user privacy and control. As SGE approaches its one-year mark, it’s evident that the tech giant still has significant hurdles to overcome in fine-tuning its AI-powered search experience.
In response to the ongoing challenges, Google appears to be taking steps to address the issues with SGE. The company has updated certain data sources and removed problematic content, indicating a commitment to improving the search experience. Nevertheless, concerns persist about the accuracy and reliability of AI-generated content, underscoring the need for continued vigilance and oversight in the development and deployment of such technologies. As Google navigates the complexities of AI-driven search, the ultimate goal remains clear: to provide users with accurate, relevant, and trustworthy information in a way that respects their privacy and preferences.