Google just broke search | Digital Trends

Google AI Overviews were announced a few weeks ago at Google I/O and have already proven to be quite controversial. The goal is to provide high-quality answers to your questions, aggregated from around the web, but a series of recent X-threads (formerly Twitter) show just how big a failure they already are.

The answer that went viral concerns a very questionable pizza recipe. As reported, when asking Google to provide an answer to the “cheese not sticking to pizza” problem, the AI ​​​​panel suggests adding non-toxic glue to the pizza to keep the cheese from sliding off. The exact words of the AI ​​​​panel are as follows: “You can also add about 1/8 cup of non-toxic glue to the sauce to make it stickier.” Where did Google's AI ​​panel source this information? From an 11-year-old Reddit comment from this thread that was clearly a joke.

pic.twitter.com/6ALCbz6EjK

– SG-r01 (@heavenrend) May 22, 2024

The words “cheese doesn't stick to pizza” prompted this unexpected and funny response, and the internet is having a blast with it. The response to the Google AI overview has since gone viral, and someone even tried sticking pizza, just to make a point.

It's worth noting that we've seen a massive increase in Reddit and forum posts appearing higher in Google searches. It's also worth noting that Reddit recently signed a $60 million deal to allow Google to train its models on Reddit content. It's not hard to see the connections as to how this might have happened.

It's not just Reddit, though. Another AI review has been posted online, with an answer to the question “How many rocks should I eat daily?” that pulls information directly from The Onion.

her pic.twitter.com/FGbvO923gk

– Tim Onion (@oneunderscore__) May 23, 2024

Part of the problem is the absolute confidence with which AI Overviews delivers its answers. It doesn't show a link to an Onion article and leave the decision up to you. Instead, it treats every source as if it were Wikipedia and delivers information with absolute confidentiality.

Google claims that its AI overviews provide users with high-quality information and that such errors are rare. Here is Google's official response to Digital Trends: “The vast majority of AI overviews provide high-quality information with links to dig deeper into the web. Many of the examples we saw were unusual queries, and we also saw examples that were manipulated or that we couldn't reproduce. We conducted extensive testing before rolling out this new experience, and as with other features we've introduced in Search, we welcome feedback. We take swift action where appropriate within our content policies and use these examples to develop broader improvements to our systems, some of which have already been rolled out.”

It seems like some of these AI overviews are going to be hard to fix if they can't be reproduced. The “broader improvements” have to be the solution, and since Google says they're already in the works, hopefully we'll see some better searches soon. We'll have to wait and see if Google responds to the situation beyond this statement. After Google received negative feedback about its image generation in Gemini earlier this year, the company had to apologize and pull it down to fix its issues.

For now, both are good reminders of how careful we need to be when trusting AI engines with information. Google AI Overview rolled out to everyone in the US earlier this month, with more countries coming soon. But with responses like these, there may be more people looking for a way to turn it off than Google anticipated.

Editor's recommendations