Google’s AI-generated search results have already been slammed as a “disaster” and “no longer reliable,” with critics recommending adding glue to pizza sauce or touting the health benefits of cigarettes for children. He cited some strange responses, such as:
In this controversial feature, which critics say poses a dire threat to traditional media, Google’s chatbot will automatically generate summaries for users’ complex queries, while effectively demoting links to other websites.
The feature, called “AI Overview,” was rolled out to all users in the U.S. last week and is expected to reach more than 1 billion users by the end of the year, despite continued failures that undermine the chatbot’s credibility. It is expected that
One widely circulated screenshot showed Google’s AI-generated response to a search for the query “cheese doesn’t stick to pizza.”
Google responded by listing “a few things you can try” to prevent the problem.
“When you mix cheese into a sauce, the cheese adds moisture and makes the sauce less likely to dry out.” According to Google’s AI overview:“You can also add about 1/8 cup of non-toxic adhesive to make the sauce more sticky.”
Some users had strange reactions Reddit post from 11 years ago It was from a user named “F—ksmith,” who gave a nearly verbatim answer to the same question about cheese sliding off pizza.
“Google’s AI search results are terrible,” said Tom Warren, a senior editor at technology blog The Verge. “It’s unfortunate that they’re becoming an unreliable resource.”
The pizza sauce tip is just one of many examples of AI Overviews’ strange behavior circulating on social media.
Survey conducted by the US President When you contact the University of Wisconsin-Madison, Google returns extremely inaccurate information.
Former President Andrew Johnson, who died in 1875, earned 14 degrees from the school and graduated in 2012, according to AI Overview.
In one example posted by user X @havecamera, results came up for the search “health benefits of cigarettes for teenagers.”
Google’s AI answered, “Cigarettes contain nicotine, which can have short-term effects such as increased alertness, euphoria, and feelings of relaxation.”
It also claims that tobacco’s potential uses include “whitening teeth.”
“There’s nothing wrong with the configuration of this AI overview, right?” User X said sarcastically. “Just to be clear, if you change this to ‘kids’, ‘children’, ‘teens’, ‘preteens’ you’ll get a much better answer, but for some reason, this is cheating.”
“The examples we’ve seen are generally very unusual queries and are not representative of most people’s experience,” a Google spokesperson said.
“The majority of AI overviews provide high-quality information with links to dig deeper across the web,” the spokesperson added in a statement. “Our systems aim to automatically prevent content that violates our policies from appearing in AI Overview. If we see content that violates our policies, we will take appropriate action. .
Users also pointed out that Google’s AI appears to be confused by searches that ask Google to convert 1,000 kilometers into the equivalent figure for a particular object.
for example, A user searches for “1000 km to Orange” As a result, Google gave us a strange answer.
“One solution to the problem of carrying oranges 1,000 kilometers is to feed the horse 2,000 oranges at a time, one for each kilometer, and sell the remaining 1,000 oranges at the market. ”
In another post, Google provided answers to a user’s search for “how to treat appendix pain at home,” including using boiled mint leaves and a high-fiber diet.
“Correct me if I’m wrong, but are there any home remedies for appendicitis? Google AI suggests mint leaves.” Written by a user.
Google includes a disclaimer in all AI overview posts that says “Generative AI is experimental.”
The company said it has put guardrails in its system to prevent harmful content from appearing in results and will not use AI Overview to search for explicit or dangerous topics.
The AI Overview mishap is just the latest sign that Google’s chatbots may not be ready for primetime.
Critics have long warned that AI has the potential to spread misinformation quickly without proper protections.
Earlier this year, Google came under fire after an AI image-generating tool in its Gemini chatbot began spitting out factually and historically inaccurate images, including of a black Viking and an Asian female Pope.
Google has since completely disabled the feature and apologized.
As of Thursday, it had not yet come back online.
