Google has recently made artificial intelligence upgrades to its search engine, but its AI Overviews feature has once again made a common-sense mistake. Multiple netizens found that when searching for the seemingly simple question "Is 2027 next year?" on Google, the AI gave a ridiculous and laughable wrong answer.

The system insisted in its answer that 2027 is still two years away from now, a shocking and basic error that quickly sparked widespread attention and discussion on social platforms.

image.png

Reusing Old Jokes and Believing Wrong Answers

After in-depth tracking, tech media found that the culprit behind the serious logical confusion of Google's AI was the "pollution" from old online posts. The system mistakenly referenced some outdated joke posts from Instagram and Reddit on social platforms when fetching answers.

These posts were originally mocking previous incorrect responses from other platforms or users on the same issue. However, Google's AI lacked the ability to distinguish humor from facts. It directly took these sarcastic wrong contents as authoritative sources and incorporated them into the final search summary.

Large Models Often Have "Blunders"

This is not the first time Google's AI search has embarrassed itself in public. Previously, the feature was criticized by the industry for giving absurd advice such as "using non-toxic glue to stick cheese on pizza." Although the official side has been continuously fixing issues, due to the fact that large models are essentially data matching based on probability, they are still easily misled by garbage information and satirical jokes on the internet.

Industry experts pointed out that this blunder once again exposed the shortcomings in current search technology's semantic understanding. Facing changing concepts of time and human-specific humorous expressions, if large models cannot establish real-time perception and logical reasoning capabilities, similar "AI hallucinations" are likely to continue happening.