Imagine this: you\u2019ve carved out an evening to unwind and decide to make a homemade pizza. You assemble your pie, throw it in the oven, and are excited to start eating.\u00a0But once you get ready to take a bite of your oily creation, you run into a problem \u2014 the cheese falls right off. Frustrated, you turn to Google for a solution.<\/p>\n<\/div>\n
\u201cAdd some glue,\u201d Google answers. \u201cMix about 1\/8 cup of Elmer\u2019s glue in with the sauce. Non-toxic glue will work.\u201d<\/p>\n<\/div>\n
So, yeah, don\u2019t do that. As of writing this, though, that\u2019s what Google\u2019s 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 \u201cfucksmith\u201d in a more than decade-old Reddit thread<\/a>, and they\u2019re clearly joking.<\/p>\n<\/div>\n This is just one of many mistakes cropping up in the new feature that Google rolled out broadly this month. It also claims that former US President James Madison graduated from the University of Wisconsin not once but 21 times<\/a>, that a dog has played in the NBA<\/a>, NFL, and NHL<\/a>, and that Batman is a cop<\/a>.<\/p>\n<\/div>\n Google spokesperson Meghann Farnsworth said the mistakes came from \u201cgenerally very uncommon queries, and aren\u2019t representative of most people\u2019s experiences.\u201d The company has taken action against violations of its policies, she said, and are using these \u201cisolated examples\u201d to continue to refine the product.<\/p>\n<\/div>\n Look, Google didn\u2019t promise this would be perfect, and it even slaps a \u201cGenerative AI is experimental\u201d label at the bottom of the AI answers. But it\u2019s clear these tools aren\u2019t ready to accurately provide information at scale.<\/p>\n<\/div>\n Take Google I\/O\u2019s big launch of this feature, for instance. The demo was highly controlled, and yet, it delivered a questionable answer<\/a> about how to fix a jammed film camera. (It suggested they \u201copen the back door and gently remove the film\u201d; don\u2019t do that unless you want to ruin your photos!)<\/p>\n<\/div>\n It\u2019s not just Google; companies like OpenAI, Meta, and Perplexity have all grappled with AI hallucinations<\/a> 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.<\/p>\n<\/div>\n 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 \u2014 boys will be boys! These companies claim that they can\u2019t predict what this AI will spit out, so really, it\u2019s out of their control.<\/p>\n<\/div>\n But for users, that\u2019s a problem. Last year, Google said that\u00a0AI was the future of search<\/a>. What\u2019s the point, though, if the search seems dumber than before?<\/p>\n<\/div>\n