Google addresses issues with its new AI Overviews after bizarre results went viral

Now, in response to the concerns, Liz Reid, VP of Google Search, acknowledged that the search engine’s AI Overviews sometimes returned “odd, inaccurate or unhelpful” results.

In a recent blog post, she addressed these issues and announced that Google has put in place safeguards to ensure the feature provides more accurate results that won’t go down in meme history.
Reid clarified that some extreme AI Overview responses, like the suggestion that it’s safe to leave dogs in cars, are fabricated. While others, like the viral screenshot depicting the response to “How many rocks should I eat?” are genuine, she clarified that Google generated the response because a website had published satirical content on the topic. She explained why the company’s AI algorithm linked to the website by saying:


She also added:

Plus, the Google VP confirmed the case where AI Overview suggested using glue to make cheese stick to pizza, citing content from a forum. While forums can offer real firsthand info, they might also give not-so-helpful tips, as you probably have noticed yourself.

Reid also mentioned that Google extensively tested the feature before launch, but “there’s nothing quite like having millions of people using the feature with many novel searches.”

Google figured out where its AI responses went wrong by looking at examples of its answers over the past few weeks. Based on what it found, the company then added safeguards. Here is what the tech giant did:

  • Google improved its ability to detect queries that don’t make sense and shouldn’t trigger an AI Overview. Additionally, the company adjusted its AI to better recognize humor and satire content.
  • The company updated its systems to reduce the reliance on user-generated content in responses that may provide misleading advice.
  • Google put in place restrictions to limit the triggering of queries where AI Overviews were not proving to be as helpful.
  • For breaking news topics, Google refrains from displaying AI Overviews to prioritize freshness and accuracy. Also, the tech giant has introduced refinements to enhance quality assurance and accuracy for health-related queries.

Source link