From Muslim Obama to Putting Glue on Pizza: Artificial Intelligence Is Poisoning Google’s Search Engine

Imagine you’re making pizza in the kitchen, but when you take it out of the oven, the cheese rolls off and doesn’t stick to the dough. You don’t know what to do to keep it from getting stuck, so you go to Google. Do you think this has all the information I need? However, when you ask seeker The new AI-generated summaries suggest you can add “about 1/8 cup of non-toxic glue to the sauce for extra stickiness.”

That’s exactly what happened this week to a user who was testing a new feature implemented by the tech giant. 10 days ago Google announced to the world that it was going to integrate AI in his popular Search enginegate to Internet for billions of people. This translates into small summaries that appear at the top of the screen containing content pulled from the Internet. “It can take a lot of hard work,” he promised. Liz Reedresponsible for this business.

However, the new search engine doesn’t work as Google expected. Over the past week and a half, dozens of users in the United States have reported that the company’s new product enhances hoaxesmakes up historical data or facts that never happened and suggests reliable information that may be dangerous, such as pizza.

“Totally dangerous”

These problems are largely due to the fact that TwinsGoogle’s AI generates summaries based on the vast amount of information it pulls from the Internet, and has no way of distinguishing between what’s true and what’s not. This means that their simplistic answers are based on unreliable sources. For example, in the case of glue and pizza, a company might accept a comment posted on a forum. Reddit 11 years ago user with the nickname “fucksmith”.

Over the past few hours, artificial intelligence experts, technology journalists and everyday users have pointed out serious problems with the search engine, from claims that only 17 of 42 US presidents have been white to pointing out that Barack Obama is a Muslim, a false conspiracy popular among the far right that has been amplified Donald Trump. “This is completely dangerous,” he complained. Melanie MitchellProfessor of Complexity at the Santa Fe Institute.

At the Google I/O conference last week, the company’s CEO Sundar Pichai, assured that this feature, while only deployed in the United States, will gradually open throughout the world. Thus, they expect that by the end of the year, more than a billion people will use these AI-generated summaries.

However, this avalanche of problems in the results AI Review may prompt the company to take action. “Could Google suspend AI reviews after a week?” asked Alex Heath, associate director of technology media at The Verge.