Google’s AI brief seems to spit out inaccurate and dangerous answers

Google’s new “AI Overview” feature artificial intelligence The ” ” that appears at the top of some search results appears to indicate inaccurate and sometimes dangerous results.

Social media users were quick to point out that the feature, which rolled out to the public earlier this month and replaced organic search results, was offering questionable answers. The engine, which has more than 90% market share in the search space, offered advice that surprised some when it came to the problem of cheese slipping off pizza: mix glue into the cheese.

But unappetizing recipes aren’t the most harmful outcome the feature can eliminate: According to an X user, in response to a prompt about disinfecting a washing machine, Google responded with a potentially deadly recipe for mustard gas (which the AI ​​feature poses a threat). Confirmed Through search.

And many large-scale language models, e.g. OpenAI Chat GPT Google’s Overview feature is pretty clear that it’s plagiarized, while Google’s search engine, Google Analytics, and Meta’s Llama search the internet for training data without permission. According to X users, Google added the phrase “my kids’ favorite” to a smoothie recipe, borrowing the phrase from a recipe website that featured similar recipes.

Google, which is developing the feature in response to Microsoft and Open AI’s Bing CoPilot, argues that the program will benefit smaller sites by allowing them to surface their content in search results and reach more searchers.

“AI Summary has led people to visit a more diverse range of websites for help with more complex questions,” the feature’s announcement states, “and links included in AI Summary have received more clicks than if the pages had been presented as traditional web listings for that query.”

Critics worry that the AI-generated results could discourage clickthroughs to the original sources Google sources, ultimately starving the publications that AI Overview relies on. The News Media Alliance, a group that represents publishers, has called for stronger measures against AI models that learn from content without permission.

“AI companies recognize the value of this content, as evidenced by marketplace arrangements, and rely on high-quality content to train their systems. It is unlawful for AI companies to continue to use publishers’ valuable content unlimitedly without proper compensation,” Alliance president and CEO Daniel Coffey said in a statement about the lawsuit against OpenAI.

Although Google includes a “Generative AI is experimental” disclosure beneath its AI-generated results, the results still pose challenges in an age of increasing reliance on computer-generated content.

If you’re looking for an alternative to tech giant Alphabet’s search engine, listen to AI. My own advice Then try a search engine that actually prioritizes what you’re looking for without bias.