Google CEO admits AI Overview feature has a "hallucination" problem

Google's recently introduced AI Overview feature, designed to provide summaries at the top of search results pages, has been criticised for spreading incorrect and strange information. In a recent interview, Google CEO Sundar Pichai acknowledged a serious problem with the new AI Overview feature, which is designed to provide brief summaries at the top of search results.

Here's What We Know

Users have encountered inaccurate and sometimes absurd information provided by the feature. Examples include advice to use glue to fix cheese on pizza and attributing false claims to experts. Pichai explained that such "hallucinations" are the result of errors in the large language models used in AI Overview. Despite the efforts of Google engineers, the problem remains unsolved, but Pichai claims that the feature can still be useful and noted an improvement in accuracy compared to last year.

Criticism from users and experts such as AI consultant Britney Mueller has emphasised the high expectations for AI accuracy and the risks associated with its premature implementation. AI Overview errors have caused discontent on the internet and could undermine the credibility of Google's search engine.

In conclusion, despite Pichai's optimism, the ongoing bugs highlight the need for further AI improvements and the importance of balancing innovation with user trust. Improving the accuracy and reliability of AI remains a priority to prevent the spread of misinformation.

Source: ITHome