Google announced its AI Overview feature at the I/O event a few weeks ago. Ever since the innovation was introduced, it has only attracted criticism due to the inaccuracy in some search results. Many users took their frustrations out on social media platforms and even turned some of the search answers into a roasting session. This bad publicity was not something the company had expected due to the rigorous internal testing before the rollout. After hearing quite the buzz around the upgrade, the company finally spoke out in its defense.
After some bizarre search results, Google is now placing more restrictions on the AI Overview
Google decided to intervene and give its side of the story after people shared odd search results and doubts about the company's vigilant approach to ensuring accuracy in its AI-powered enhancements. Liz Reid, Head of Google Search, shared the errors, the causes, and the steps for the rectification in a blog post.
They’re not simply generating an output based on training data. While AI Overviews are powered by a customized language model, the model is integrated with our core web ranking systems and designed to carry out traditional “search” tasks, like identifying relevant, high-quality results from our index. That’s why AI Overviews don’t just provide text output, but include relevant links so people can explore further.
The company was in denial regarding the AI Overview hallucination and suggested that it does not work in the same capacity as other LLM models as it is supported by the most relevant web search results. It presented certain possibilities where the feature could present incorrect information, which was related to either misunderstanding the queries or having sparse information on the source.
Guarding their product image further after the unpopular "glue-on-pizza search result," Google blamed some users for their illogical searches that generated controversy and for faking some of the Overviews. It did admit the information gaps in the search results that do not interpret satire quite well. The company also stated its efforts to ensure the problem is tackled on a macro level rather than going query-to-query.
Google shared some of the changes it made to improve the overall AI Search experience, including filtering for satire and illogical searches, limiting UGC in responses, and restricting AI Overviews in entirety for news and health. Even when Google emphasized the testing it went through and the high standard it maintains, the company continued highlighting its determination to ensure AI Overview precision. Have you had any odd experiences with AI Overview? Share with us in the comments below.
Follow Wccftech on Google to get more of our news coverage in your feeds.
