Last week, Google unveiled AI Overview, the company’s latest technology that is intended to help answer people’s questions quickly. However, social media users have pointed out that the new technology is generating incorrect answers instead, including recommending glue as part of a pizza recipe or ingesting rocks to boost your immune system.
Google describes AI Overview — its biggest change to its search engine in years — as “letting Google do the Googling for you.” The new technology pulls snippets from sites around the web to generate an answer. It can even cite its sources but doesn’t know when that source is incorrect — or in the case of an article from The Onion, is being humorous.
“Google invented the transformer they will win the AI race”
Google AI: pic.twitter.com/dLquFIBRWS
— Dr. Parik Patel, BA, CFA, ACCA Esq. (@ParikPatelCFA) May 24, 2024
Google shared the system was designed to answer more complex and specific questions than regular search. But, like most things on the Internet, things went awry, and users posted screenshots of baffling responses to social media platforms like X.
As long as it’s non-toxic pic.twitter.com/VIv6ozVwQN
— Jeremiah Johnson 🌐 (@JeremiahDJohns) May 26, 2024
“smoking while pregnant” pic.twitter.com/egFgmupYvW
— Jeremiah Johnson 🌐 (@JeremiahDJohns) May 26, 2024
“train eight days a week” pic.twitter.com/xlDkOQEKo8
— Jeremiah Johnson 🌐 (@JeremiahDJohns) May 26, 2024
Looney Tunes moment pic.twitter.com/dzqgv55J0F
— Jeremiah Johnson 🌐 (@JeremiahDJohns) May 26, 2024
“how long can I stare at the sun” pic.twitter.com/Q22V490f6F
— Jeremiah Johnson 🌐 (@JeremiahDJohns) May 26, 2024
This isn’t the first AI chatbot to tell lies and act weird, and it’s not the first time for Google’s AI algorithm to get things wrong. In February 2023, Google launched Bard, a chatbot to rival ChatGPT, which reportedly shared incorrect information about a wide variety of topics. (The company’s market value subsequently dropped by $100 billion.)
This February, Google released Gemini, a chatbot that could generate images and act as a voice-operated digital assistant. Users quickly pointed out that the system refused to generate images of white people in many instances and drew inaccurate depictions of historical figures.
Google has shared that they are working to fix the problem, but industry experts think it might be too late. As one AI founder told The Verge, “A company once known for being at the cutting edge and shipping high-quality stuff is now known for low-quality output that’s getting meme’d.”
