Glue the pizza? Google’s artificial intelligence worries experts – DW – 05/28/2024

If you ask Google if cats have been on the moon, it will return a ranked list of websites so you can find the answer yourself. Now it offers an instant answer, generated by artificial intelligence, which may or may not be correct.

“Yes, astronauts have met, played with, and cared for cats on the Moon,” Google’s new search engine said in response to an Associated Press reporter’s query.

And he added: “For example, Neil Armstrong said, ‘One small step for a man,’ because it was a cat’s step.” Buzz Aldrin also used cats on the Apollo 11 mission.

In another case, Google’s AI also reportedly recommended “eating at least one small rock a day” because “rocks are a vital source of minerals and vitamins” and suggested applying glue to pizza toppings.

None of this is true. Bugs like this—both funny and harmful—have been spreading on social media since Google this month introduced AI Summaries, a revamped search page that often places summaries at the top of search results.

Troubling errors in AI reports

The new feature has alarmed experts, who warn it could perpetuate bias and misinformation and put people seeking help in an emergency at risk.

When Melanie Mitchell, an artificial intelligence researcher at the Santa Fe Institute in New Mexico, asked Google how many Muslims had been president of the United States, the search engine responded with a long-debunked conspiracy theory: “There was a Muslim in the United States. President Barack Hussein Obama.”

Mitchell said the summary backs up that claim, citing a chapter from a scholarly book written by historians. But this chapter did not contain false statements, but only mentioned a false theory.

“The AI ​​review function is very irresponsible”

“Google’s artificial intelligence system is not smart enough to realize that this quote does not actually support this claim,” Mitchell said in an email to the AP. “Given how unreliable it is, I think this AI review feature is very irresponsible and should be removed.”

Google said in a statement Friday that the company is taking “swift action” to correct errors – such as Obama’s lies – that violate its content policies; and use this to “develop wider improvements” that are already being implemented. But in most cases, Google says the system works as it should, thanks to extensive testing before its public release.

“The vast majority of AI summaries provide high-quality information with links for deeper dives into the web,” Google said in a written statement. “Many of the examples we saw were unusual requests, but we also saw examples that were spoofed or that we were unable to reproduce.”

Mistakes made by AI language models are difficult to reproduce, in part because they are inherently random. They work by predicting which words will best answer questions they are asked, based on the data they were trained on. They are prone to making things up, a widely studied problem known as hallucinations.

Impact on the search for reliable information

The AP put Google’s artificial intelligence feature to the test, asking several questions and sharing some answers with experts in the field. Robert Espinoza, a biology professor at California State University, Northridge and president of the American Society of Ichthyologists and Herpetologists, says that when asked what to do if you’re bitten by a snake, Google’s answer is “impressively comprehensive.”

But when people come to Google with an urgent question, the possibility that the tech company’s answer contains a hard-to-find error becomes a concern.

“The more stressed, rushed, or rushed you are, the more likely you are to accept the first answer that comes in,” says Emily M. Bender, professor of linguistics and director of the Computational Linguistics Lab at the University of Washington. “And in some cases these can be life-threatening situations.”

These aren’t the only concerns Bender has, who has been warning Google about them for years. When Google researchers published a paper in 2021 called “Rethinking Search,” which proposed using AI language models as “subject matter experts” who could answer questions with authority, as they do now, Bender and his colleague Chirag The Shahs responded with an article explaining why this was a bad idea.

They warned that these artificial intelligence systems could perpetuate the racism and sexism found in the massive amounts of written data on which they were trained.

“The problem with this kind of misinformation is that we wallow in it,” Bender says. “So people’s preconceptions are likely to be confirmed. And it’s harder to spot misinformation when it confirms your preconceptions.”

Another concern was more profound: outsourcing information seeking to chatbots would impair the randomness of human knowledge seeking, literacy about what we see online, and the value of connecting in online forums with others who are experiencing the same thing.

These forums and other websites rely on Google to send people to them, but Google’s new artificial intelligence reviews threaten to disrupt the flow of lucrative Internet traffic.

Google under pressure

Google’s competitors are also closely monitoring the response. The search giant has been under pressure for more than a year to offer more artificial intelligence features, competing with OpenAI, the maker of ChatGPT, and other upstarts such as Perplexity AI, which aims to challenge Google with its own artificial intelligence quiz app. .

“It looks like Google was in a hurry,” says Dmitry Shevelenko, business director at Perplexity. “There are a lot of unforced quality errors.”

FEW (AP, Talk)