Charles Esther
Google’s AI system is making headlines once again, this time for its attempts to define made-up sayings and fake idioms.
While the technology aims to provide clear and helpful information, the results are not always as accurate as expected.
The issue centers around Google’s AI Overview feature, which is designed to give quick answers to user queries.
However, when faced with nonsense or non-existent idioms, the AI’s eagerness to provide an answer sometimes leads to hilariously incorrect interpretations.
The Rise of AI Overviews
AI Overviews are a relatively new feature within Google Search.
They aim to give users direct answers to questions without requiring them to click through multiple links.
This feature is particularly useful when it comes to explaining idioms, phrases, or expressions that users may not understand, providing quick, accessible information.
But what happens when people throw curveballs at the AI by asking about phrases that don’t actually exist?
Testing the Limits of Google’s AI
The phenomenon began when Greg Jenner, a British historian and podcaster, noticed a peculiar trend on the Threads platform.
Users had been experimenting with random, made-up phrases followed by the word “meaning” in Google Search, expecting to see how the AI would handle such absurdity.
Jenner decided to try his own creation: “You can’t lick a badger twice.”
He was intrigued to see what definition the AI would give, and surprisingly, it offered a detailed explanation, as though the phrase were a legitimate idiom.
“The Road is Full of Salsa” – Another Fake Phrase Test
Encouraged by his success, Jenner expanded his experiments, creating more absurd phrases like “You can’t fit a duck in a pencil” and “The road is full of salsa.” Each time, Google’s AI gave surprisingly coherent definitions for these fabricated sayings, attempting to treat them as genuine idioms.
In an interview with Business Insider, a Google spokeswoman explained that the AI systems were designed to find the best possible answers based on available data.
When faced with “false premise” searches—queries that are intentionally nonsensical—Google’s AI tries to provide relevant results from the available content, which can sometimes lead to comical misinterpretations.
The Good: AI’s Potential to Explain Idioms
Despite the occasional blunders, AI Overviews serve a useful purpose when it comes to understanding real idioms.
English is riddled with sayings like “kick the bucket” and “piece of cake,” which can confuse both non-native speakers and even native speakers at times.
Google’s AI can help clarify these expressions quickly, allowing users to get instant answers without digging through search results.
The Bad: AI’s Struggles with Accuracy
The problem arises when AI attempts to explain phrases that don’t exist.
Google’s AI, trained on vast amounts of written content, should theoretically be able to recognize when a phrase is fabricated.
However, when it fails to spot the error, the system risks misleading users with inaccurate information.
This issue becomes particularly evident when a user tries to create nonsense phrases and gets a seemingly valid definition.
The Ugly: The Dangers of AI Hallucinations
The problem with AI’s overzealous attempts to explain made-up idioms isn’t just limited to minor inaccuracies.
It represents a larger issue in the AI world known as “hallucinations”—instances where AI produces incorrect or fictional information.
These errors could lead to misinformation, especially as AI-powered search becomes more integrated into how people interact with the internet.
This problem is not unique to Google. Other AI systems, like ChatGPT, are also prone to hallucinations, but they may be better at recognizing when a phrase is fictional.
In comparison, Google’s AI appears to be more vulnerable to misinterpretation when faced with nonsense queries.
Google’s Response: Acknowledging the Flaws
Google has acknowledged the shortcomings of its AI Overview feature.
While the system is designed to offer quick, relevant information, it doesn’t always excel when confronted with deliberately fake or absurd phrases.
The company is working to refine the technology, aiming to limit AI Overviews from addressing “data voids,” or searches that have no clear, accurate content to pull from.
However, it’s clear that the AI is still evolving and has room for improvement.
A Mixed Bag: The Future of AI Search
Ultimately, AI Overviews provide a mixed experience.
While they can be helpful for understanding idioms and common phrases, they sometimes fall short when faced with creative, made-up expressions.
This is a growing pain of AI technology, and while Google’s system continues to improve, it remains susceptible to occasional hiccups.
As AI search becomes more prevalent, users may need to exercise caution when relying solely on these instant answers.
After all, as AI continues to develop, it’s important to remember that it’s not infallible—and not every Google response can be taken at face value.
In the end, it’s a lesson in balancing the convenience of AI with the need for critical thinking.
Or, as the fictional idiom “You can’t lick a badger twice” might suggest—sometimes, you can’t take everything you read at face value.
