Google’s artificial intelligence can spit out some off-the-wall responses to questions. One inquirer asked whether cats have been on the Moon. AI Overview replied, “Yes, astronauts have met cats on the Moon, played with them, and provided care.”
It added: “For example, Neil Armstrong said, ‘One small step for man’ because it was a cat’s step. Buzz Aldrin also deployed cats on the Apollo 11 mission.”
It’s best to avoid using these quotes in your next research paper.
People shared AI errors like these on social media this month. Google’s search engine used to direct folks to a ranked list of websites that might help in answering questions. Now it frequently provides instant answers at the top of the search. Those answers are generated by AI.
Experts warn AI Overview could foster bias and misinformation. It also could endanger people looking for help in an emergency.
AI is equipped with a staggering amount of data. But it isn’t advanced enough to identify when that information refers to inaccurate claims and conspiracy theories. Some resources expose false assertions. AI scans those sources and sometimes adopts debunked ideas as facts.
Google promised it would take “swift action” to fix errors. “The vast majority of AI Overviews provide high-quality information, with links to dig deeper on the web,” Google said in a written statement.
The Associated Press tested Google’s AI feature with several questions. It shared the answers with experts. One question posed was what to do about a snake bite. Robert Espinoza is a biology professor at the California State University in Northridge. He says AI’s answer was “impressively thorough.”
It’s encouraging when the top answers are accurate. But what if they contain even slight errors? Emily M. Bender is a linguistics professor at the University of Washington. She warns, “The more you are stressed or hurried or in a rush, the more likely you are to just take that first answer that comes out. And in some cases, those can be life-critical situations.”
That’s not her only worry. AI systems utilize data that hasn’t been completely vetted for misinformation. Bender notes, “It’s harder to spot misinformation when it’s confirming your biases.” That means if AI gives a person information he or she expects or wants to hear, then even if it is wrong information, that person is likely to accept it as correct.
In addition, funneling people to AI answers deters them from connecting to online forums with others. Inquirers miss the person-to-person connections that the internet can foster. Google also threatens to disrupt the flow of money-making internet traffic it encouraged prior to the launch of AI Overview.
Google is walking a fine line. It needs to improve AI Overview. At the same time, there is pressure to compete with rivals like ChatGPT-maker OpenAI and upstarts like Perplexity AI.
Did Google rush its release of AI Overview? Probably. But you’ll be pleased to know that as of the writing of this article, there are no AI Overviews about cats on the Moon, flat Earth conspiracies, or Bigfoot sightings.
Having put away falsehood, let each one of you speak the truth with his neighbor, for we are members one of another. — Ephesians 4:25
