Google, the tech giant owned by Alphabet Inc., recently introduced the AI Overviews feature to users in the United States.
However, after nearly two weeks of its rollout, the feature has been met with a barrage of complaints and concerns from users who have shared screenshots of the AI’s unusual and often inaccurate responses on social media platforms like X (formerly Twitter) and Threads.
One particularly troubling example occurred when a user sought advice on preventing cheese from falling off their pizza. Google’s AI, based on a Reddit comment, suggested adding glue to the sauce, failing to detect the sarcasm in the original post.
In another instance, the AI claimed that former U.S. President Barack Obama was the nation’s first Muslim President.
Meghann Farnsworth, a Google spokesperson, acknowledged these errors, attributing them to “generally very uncommon queries.”
Prediction time: Google AI overviews will be disabled in the next 15 days, similar to the image generation of people in Gemini. It's not working as expected, and here are some hilarious unexpected response 🧵
1/n pic.twitter.com/x0w6nHaP50— AshutoshShrivastava (@ai_for_success) May 24, 2024
She assured users that the company is taking action against policy violations and using these “isolated examples” to improve the product.
However, the widespread deployment of the AI Overviews feature, despite its evident flaws, underscores the challenges faced by companies developing AI technologies.
TRENDING: Legal Filing Shows Why NO ONE Should Ever Fly On American Airlines
Google has admitted that the feature is not perfect, labeling it as “Generative AI is experimental” at the bottom of the AI-generated answers.
Nevertheless, it is clear that these tools are not yet capable of accurately providing information at scale.
This issue is further highlighted by Google Trends data, which indicates that many users are searching for ways to disable the AI feature on Google, despite CEO Sundar Pichai’s promotion of its success.
Elon Musk on Google Gemini's wokeness: "I think that's extremely dangerous because the AI could conclude the best way to avoid misgendering would be to destroy all humans." pic.twitter.com/AXhTldS7rK
— Ian Miles Cheong (@stillgray) May 23, 2024
This is not the first time Google has faced criticism over its AI innovations.
In February 2023, the company’s Gemini AI chatbot came under fire for generating historically inaccurate images and text.
The issue was deemed “completely unacceptable” by Pichai, and Google announced plans to pause the image generation of people and release an improved version in the future.
As AI continues to advance and integrate into various aspects of our lives, it is crucial for companies like Google to prioritize accuracy, reliability, and user experience.