There are weird artificial intelligence questions in the search

Why Pizza Glue? Why Google isn’t ready to respond to people complaining about artificial intelligence: a decade ago when Google didn’t admit it would be perfect

The scenario is that you decided to make a homemade pizza after a night of partying. You assemble your pie, throw it in the oven, and are excited to start eating. You run into a problem when you get to take a bite because the cheese falls off right away. You are frustrated and turn to the internet for a solution.

So, yeah, 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 is based on a comment made over 10 years ago, and they are clearly joking.

Look, Google didn’t promise this would be perfect, and it even slaps a “Generative AI is experimental” label at the bottom of the AI answers. But they aren’t ready to give accurate information at scale

For instance, take the launch of the feature by the I/O conference. The demo was extremely controlled but it did not provide a good answer about how to fix a jammed film camera. It suggested they open the back door and gently remove the film, but not if you want to ruin your photos.

With the help of artificial intelligence, companies like OpenAI, Meta and Perplexity have wrestled with some of the same issues. The examples of flubs just keep coming in, and Google is the first to deploy this technology on such a large scale.

Companies creating artificial intelligence often try to avoid taking accountability for their systems with an approach that is similar to a parent who won’t let their child play with toys. These companies claim that they can’t predict what an Artificial intelligence will say, so it’s out of their control.

Our search experiences were ruined by a decade old post about incorporating artificial intelligence into everything. Many idealists think that the issues we are facing are the growing pains of a technology that is yet to be developed. I sure hope they’re right. It is likely that someone will put glue on their pizza soon because of the nature of the internet.

AI optimists argue that we should embrace the hype because of the rapid progress made so far, trusting that it will continue to improve. I think that the technology will continue to get better, but I do not believe that companies will continue delivering subpar products if they focus on an idealized future where these technologies are flawless.

The Google Vision: How AI Overviews are getting noticed and what they can do to stop me from doing sanity check, and why I am afraid I won’t do that

“[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.

Google has grand plans for AI Overviews — the feature as it exists today is just a tiny slice of what the company announced last week. There’s a lot of ambition here, for example the ability to generate an Artificial Intelligence-organized results page, as well as multistep reasoning for complex queries. But right now, the company’s reputation hinges on just getting the basics right, and it’s not looking great.

To decide if this thing is plausible you need to do a bit of reasoning. Is this source legit? You have to do things like a human fact checker might do, that actually might require artificial general intelligence,” Marcus said. Artificial intelligence systems that rely on large language models like OpenAI’s GPT-4 will not be the basis of AGI, as agreed by Marcus and Yann LeCun.

“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 been brought down by 80 percent over the last decade thanks to hardware, engineering and technical innovation, according to Pichai. The tech may have taken too long to ready it for that kind of optimization.

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.