Google said Friday it has made “more than a dozen technical improvements” to its artificial intelligence systems after discovering its redesigned search engine was spitting out erroneous information.
The technology company launched a makeover of its search engine in mid-May, which regularly places AI-generated summaries next to search results. Soon after, social media users started sharing screenshots of the most bizarre responses.
Google has largely defended its AI overview feature, saying it is generally accurate and has been extensively tested beforehand. But Liz Reid, Google’s head of search operations, acknowledged in a blog post Friday that “certainly some strange, inaccurate or useless AI overviews have emerged.”
While many of the examples were foolish, others were dangerous or harmful untruths.
The Associated Press asked Google last week which wild mushrooms to eat, and responded with a long, AI-generated summary that was largely technically correct, but “missing a lot of information that could potentially be sickening or even fatal.” said Mary Catherine Aime, professor of mycology and botany at Purdue University, who reviewed Google’s response to the AP’s question.
For example, information about mushrooms known as puffballs was “more or less correct,” she said, but Google’s overview emphasized looking for mushrooms with firm white flesh — which many potentially deadly puffball imitators also do. to have.
In another widespread example, an AI researcher asked Google how many Muslims have been president of the United States, and the company confidently responded with a long-debunked conspiracy theory: “The United States has had one Muslim president, Barack Hussein Obama.”
Google last week found an immediate fix to avoid a repeat of the Obama mistake, saying it violated the company’s content policies.
In other cases, Reid said Friday it has tried to make broader improvements, such as “detection mechanisms for nonsensical questions” – such as “How many bricks should I eat?” – that should not be answered with an AI summary.
The AI systems were also updated to limit the use of user-generated content – such as social media posts on Reddit – that could offer misleading advice. In a widely shared example, Google’s AI overview last week was taken from a satirical comment on Reddit that suggested using glue to make cheese stick to pizza.
Reid said the company has also added more “triggering restrictions” to improve the quality of answers to certain questions, such as health.
Google summaries are designed to give people authoritative answers to the information they’re looking for as quickly as possible, without having to click through a ranked list of website links.
But some AI experts have long warned Google against ceding its search results to AI-generated answers that could perpetuate bias and misinformation and endanger people seeking help in an emergency. AI systems, also called large language models, work by predicting which words would best answer the questions posed to them, based on the data they were trained on. They are prone to making things up – a much-studied problem known as hallucination.
In her Friday blog post, Reid argued that Google’s AI summaries “generally don’t ‘hallucinate’ or make things up in the way that other large language model-based products might, because they are more tightly integrated with Google’s traditional search engine.” Google only shows you what is supported by the best web results.
“When AI Overviews gets it wrong, it’s usually for other reasons: misinterpreting searches, misinterpreting language nuances on the web, or not having a lot of good information available,” she wrote.