Google Gives An Explanation Regarding Strange Suggestions From AI Overview

JAKARTA AI Overview, a summary feature based on Artificial Intelligence (AI) on Google Search, has received criticism since May 24. This new feature has been heavily criticized for providing misinformation and misleading. Google finally provided an explanation regarding this issue. Initially, Google Search Head Liz Reid, said that AI Overview was designed to make it easier for Search users. The reason is, this feature can provide answers to complex questions. Despite being powered by AI, this feature is different from chatbots or Large Language Models (LLMs) otherwise because it is integrated into the core of the system from web rankings. With this integration, AI Overview can identify relevant and high-quality information. "That's why AI Overview not only provides text output, but includes relevant links so that people can explore further," Reid said. "AI Overview is made to display only information supported by the top web results." Reid explains that AI Overview is notACTS like other LLMs, but unreasonable tracing can occur due to several reasons. One of Reid's summary results mentions is the use of glue to compactify cheese into pizza. From the team's exploration results, it is found that AI Overview cites sarcastic content from discussion forums. This forum often shares original and good information, but other discussions can be less useful. AI Overview also misinterprets the language.

"We see the AI summary misinterpret the language on the web page and present inaccurate information. We are working quickly to address this issue, either through the refinement of our algorithm or through a process that has been set to remove responses that do not comply with our policies," Reid said. This strange and inaccurate response is being addressed by the Google Search team. So far, the company is improving AI Overview detection capabilities, updating the system to limit content usage, and adding trigger constraints. "In addition to this refinement, we are also very vigilant in monitoring external inputs and reports, as well as taking action against a small number of AI certificates that violate content policies," Reid explained.