How to Fix the Cheese Falloff: Google’s AI Overviews Explains the Pizza Glue Query in an Unusual Environment

Imagine this, you carve out an evening to relax and make a homemade pizza. You assemble your pie, throw it in the oven, and are ready to eat. But once you get ready to take a bite of your oily creation, you run into a problem — the cheese falls right off. Frustrated, you turn to Google for a solution.

Don’t do that. As of writing this, though, that’s what Google’s new AI Overviews feature will tell you to do. The feature, while not triggered for every query, scans the web and drums up an AI-generated response. The answer received for the pizza glue query appears to be based on a comment from a user named “fucksmith” in a more than decade-old Reddit thread, and they’re clearly joking.

The product outputs high quality information to users, according to the company. Many of the examples we have seen have been uncommon queries and we also saw examples that were doctored or unable to reproduce, according to a spokeswoman for the search engine. In order to develop larger improvements to our systems, the company is taking swift action to remove the AI Overviews on certain queries where appropriate under our policy, as well as using the examples to develop larger improvements.

Take Google I/O’s big launch of this feature, for instance. The demo was highly controlled, and yet it delivered a questionable answer to fix a jammed film camera. They suggested to open the back door and gently remove the film, unless you want them to ruin your photos.

It’s not just Google; companies like OpenAI, Meta, and Perplexity have all grappled with AI hallucinations and mistakes. However, Google is the first to deploy this technology on such a large scale, and the examples of flubs just keep rolling in.

Companies developing artificial intelligence are often quick to avoid taking accountability for their systems with an approach much like a parent with an unruly child — boys will be boys! These companies claim that they can’t predict what this AI will spit out, so really, it’s out of their control.

That is a problem for users. Last year, Google said that AI was the future of search. If the search seems dumber than before, what is the point?

Gary Marcus, an AI expert and an emeritus professor of neural science at New York University, told The Verge that a lot of AI companies are “selling dreams” that this tech will go from 80 percent correct to 100 percent. Marcus said that it is easy to achieve 80 percent and difficult to achieve the final 20 percent. Marcus thinks that the hardest thing to do is the last 20 percent.

“[These models] are constitutionally incapable of doing sanity checking on their own work, and that’s what’s come to bite this industry in the behind,” Marcus said.

Getting AGI Right Before Memo’d: Why Google Hasn’t Solved Its Expected Potentially Million Dollars?

Today’s feature is just part of the larger plan, which was announced last week. Multistep reasoning for complex queries, the ability to generate an AI-organized results page, video search in Google Lens — there’s a lot of ambition here. But right now, the company’s reputation hinges on just getting the basics right, and it’s not looking great.

It’s necessary to do some logic to decide if this is plausible. Is this source legit? Marcus said artificial general intelligence may need to be used to do things like a human fact checker. And Marcus and Meta’s AI chief Yann LeCun both agree that the large language models that power current AI systems like Google’s Gemini and OpenAI’s GPT-4 will not be what creates AGI.

“A company once known for being at the cutting edge and shipping high-quality stuff is now known for low-quality output that’s getting meme’d,” one AI founder, who wished to remain anonymous, told The Verge.

The cost of delivering artificial intelligence answers has fallen by 80% over the course of a decade, as a result of hardware, engineering and technical breakthrough bygoogle, according to Pichai. It appears that kind of optimization might have happened too early, before the tech was ready.

Social media is abuzz with examples of Google’s new AI Overview product saying weird stuff, from telling users to put glue on their pizza to suggesting they eat rocks. The messy rollout means Google is racing to manually disable AI Overviews for specific searches as various memes get posted, which is why users are seeing so many of them disappear shortly after being posted to social networks.