Google’s new AI search feature has sparked concern among netizens due to several inaccurate responses, such as suggesting the use of “non-toxic glue” in pizza preparation, eating rocks, and using chlorine gas to clean washing machines. Most notably, when a user sought help for depression, the AI recommended jumping off the Golden Gate Bridge, prompting severe criticism and viral backlash.
In response, Google defended these errors as “isolated examples,” claiming they were uncommon queries and not representative of most users’ experiences. However, the root cause of these inaccuracies remains unclear, leaving doubts about Google’s future in the AI domain.
“The examples we’ve seen are generally very uncommon queries, and aren’t representative of most people’s experiences. The vast majority of AI overviews provide high-quality information, with links to dig deeper on the web,” the company stated.
One potential explanation lies in the inherent tendency of language models like Google’s AI and OpenAI’s GPT-4 to “hallucinate,” generating false information without warning. Google CEO Sundar Pichai acknowledged this issue, stating that finding a solution to these hallucinations remains unresolved.
This incident isn’t Google’s first AI blunder of the year. Earlier, their Gemini AI came under fire for displaying biased and inaccurate images, such as Black Vikings and racially diverse Nazi soldiers. Google’s apology cited a failure in tuning to ensure a diverse range of images and an overly cautious model that refused to answer certain prompts entirely.
Prabhakar Raghavan, Google’s senior vice president of knowledge and information, explained that Gemini’s errors stemmed from a failure to anticipate scenarios where a range of images wasn’t appropriate and from the model becoming excessively cautious over time.
“So what went wrong? In short, two things. First, our tuning to ensure that Gemini showed a range of people failed to account for cases that should clearly not show a range. And second, over time, the model became way more cautious than we intended and refused to answer certain prompts entirely,” he explained.
These incidents highlight the challenges and risks associated with deploying AI at scale, particularly in critical areas like search. While AI holds immense potential to enhance user experiences and streamline processes, its susceptibility to errors and biases necessitates careful development and ongoing refinement.
As Google navigates these challenges, addressing issues of accuracy, bias, and user trust will be paramount. Transparency in AI development, robust testing protocols, and proactive measures to mitigate biases can help restore confidence in AI systems and ensure they deliver reliable and relevant information to users.