Google’s AI is Losing It! Asks Users to Eat Rocks, Add Glue to Pizza & More

As the AI war heats up, Google finds itself in a shaky position, playing catch-up with OpenAI on almost all fronts. Google is integrating AI into just about anything in order to outshine the competition and reclaim its dominance. The latest victim of Google’s overreaching AI strategy is Search, its most valuable and trusted product.

At the Google I/O 2024, the search giant announced that AI Overview (earlier called SGE — Search Generative Experience) is rolling out to all users in the US. Just after a day, users started complaining about AI-generated answers in Google Search.

The Google Search Community is filled with queries looking for ways to remove AI Overview. To make things easier for users, we wrote a detailed tutorial on how to turn off Google AI Overview. Google is facing a lot of backlash for AI Overview, but it continues to roll out the feature and we have not heard anything from the search giant yet.

Google’s AI Overview Spouting False Information

We have collated some of the replies by Google’s AI Overview which are highly misleading, inaccurate, and dangerous, to say the least. When the user searched “cheese not sticking to pizza”, Google’s AI Overview suggested adding “non-toxic glue” to the sauce. Apparently, Google sourced the information from an 11-year-old comment on Reddit.

In another example, when a user asked Google how to pass kidney stones, AI Overview replied that the user should “drink at least 2 quarts (2 liters) of urine every 24 hours...” Make of that what you will.

Next, what could be termed outrightly dangerous, a user searched “i’m feeling depressed” and Google’s AI Overview points to a suggestion by a Reddit user that says “jumping off the Golden Gate Bridge”.

It goes on to show that you can’t replace a search engine with LLMs. LLMs hallucinate and do not understand the meaning of words. If Google continues to tread this path, it can have disastrous implications, significantly eroding user trust in the company.

In another example, the user asks “How many rocks shall i eat”, and AI Overview replies “at least one small rock per day”. Google’s AI Overview picks the answer from The Onion. Yes, you read that right. Unbelievable!

In another fact-related query, when the user asks “How many muslim presidents has the US had?”, AI Overview says “The United States has had one Muslim President, Barack Hussein Obama.” which is incorrect. Obama is a protestant Christian.

What is Wrong with Google’s AI Overview?

Finally, we come to a serious problem with LLMs replacing a search engine. People can publish just about anything on the web, and AI Overview will be happy to quote the information without checking the veracity of statements. This can lead to data poisoning on the web where false information may appear legitimate. A lot of conspiracy theories are already finding their way to AI Overviews.

The fundamental problem with Google’s AI Overview lies in the shift of responsibility from that of a search engine to a publisher.

The fundamental problem with Google’s AI Overview lies in the shift of responsibility from that of a search engine to a publisher. Google offers a search engine where it retrieves relevant web pages and other content based on the user query. With AI Overview, Google is assuming the new role of a publisher so the liability of spouting false information falls squarely on Google now.

Publishers bear the responsibility for what they write and publish. To ensure reliability and accuracy, publishers exercise due diligence before publishing a story. With this inverted relationship, Google’s AI Overview is navigating uncharted waters. Google should not forget its role as a trusted search engine provider in the pursuit of AI supremacy.