TECH NEWS – The search engine feature unveiled at I/O has already produced some very interesting results (we’ve already reported on it), but Google is defending the AI overview…

The Alphabet subsidiary did not expect such negative feedback on the feature, as it has been thoroughly tested internally. In the first round, Google only said that unusual queries caused AI Overview to give strange answers (although there wasn’t much truth in that), but now the company has defended its search engine innovation in a more specific response.

Google’s head of search, Liz Reid, wrote in a blog post: “They don’t simply generate a result based on training data. While AI Overviews are powered by a custom language model, that model is integrated with our core web ranking systems and designed to perform traditional “search” tasks, such as identifying relevant, high-quality results from our index. That’s why AI Overviews don’t just provide text output, but also include relevant links for people to explore further.”

Google denies that AI overviews are hallucinatory and do not work in the same capacity as other large language models (LLMs) because they rely on the most relevant search results. There are several ways in which the feature could return incorrect information, either due to misunderstanding the query or lack of information. Google has blamed some users for illogical searches (such as the “glue in pizza sauce” result) and says that some AI overview results have been faked, but has admitted that it cannot interpret the satire well enough due to information gaps in the search results. They will handle cases on a macro level, not query by query.

Google has also shared some changes to the AI search experience: filtering out satire and illogical searches, limiting user-generated content in responses, and limiting AI Overview to news and health searches only. The question is whether this will convince users…

Source: WCCFTech, Google Blog