Recently, Google has made swift action to address inaccurate results generated by its AI Overview feature powered by Gemini. The tool, introduced last year as a flexible AI that can run in data centers and mobile devices, offered in three different sizes (Ultra, Pro, and Nano). Users have shared on social media incorrect answers generated by the feature. For example, one user shared a recommendation from geologists at the University of Berkeley to eat at least one small stone a day. Another user mentioned that the tool recommended using non-toxic glue on a pizza to help the cheese melt with the dough.
Google has stated that most responses generated by the tool are of high quality, but these inaccuracies are likely due to uncommon queries or possible manipulation. To improve the accuracy of its systems, Google plans to eliminate general descriptions of AI in certain queries when appropriate. It also aims to eliminate responses that include biased or erroneous information, known as hallucinations. The company is working quickly to address these issues and provide users with reliable and accurate information through its AI Overview feature.
