Google promised better search – now it’s telling us to put glue on pizza

Picture this: you’ve booked a night out to relax and decide to make homemade pizza. You assemble your pie, throw it in the oven, and can’t wait to start eating. But once you’re ready to take a bite of your oily creation, you run into a problem: the cheese immediately falls out. Frustrated, you turn to Google to find a solution.

“Add glue,” Google responds. “Mix about 1/8 cup of Elmer’s Glue with the sauce. A non-toxic glue will do the trick.

So yeah, don’t do that. As of this writing, however, that’s what Google’s new AI Overviews feature will tell you to do. The feature, while not triggered for every request, scans the web and generates an AI-generated response. The response received for the pizza glue query appears to be based on a comment from a user named “fucksmith” in a decade-old Reddit thread, and they are clearly joking.

This is just one of many mistakes made in the new feature that Google rolled out widely this month. He also claims that former U.S. President James Madison graduated from the University of Wisconsin not once but 21 times, that a dog played in the NBA, NFL and NHL and that Batman is a cop.

Look, Google hasn’t promised it’ll be perfect, and it even puts a “Generative AI is experimental” label at the bottom of the AI ​​responses. But it is clear that these tools are not ready to provide precise and large-scale information.

Take for example the big launch of this feature by Google I/O. The demo was highly controlled, and yet it provided a questionable answer on how to fix a stuck camera. (He suggested they “open the back door and gently remove the film”; don’t do this unless you want to ruin your photos!)

It’s not just about Google; companies like OpenAI, Meta, and Perplexity are all struggling with AI hallucinations and errors. However, Google is the first to deploy this technology on such a large scale, and the examples of errors keep coming.

Companies developing artificial intelligence often avoid taking responsibility for their systems, taking an approach similar to that of a parent with an unruly child: boys will be boys! These companies claim they can’t predict what this AI will produce, so in reality it’s out of their control.

But for users, it’s a problem. Last year, Google declared that AI was the future of search. What’s the point, though, if the research seems dumber than before?

AI optimists argue that we should embrace this hype because of the rapid progress made so far, and be confident that it will continue to improve. I truly believe that this technology will continue to improve, but focusing on an idealized future in which these technologies will be flawless ignores the significant problems they currently face and allows companies to continue providing substandard products.

For now, our search experiences are marred by decade-old Reddit posts with the aim of integrating AI into everything. Many idealists believe that we are on the cusp of something great and that these problems are just the growing pains of a nascent technology. I sure hope they’re right. But one thing’s for sure: We’ll probably soon see someone putting glue on their pizza, because that’s the nature of the Internet.