© IE Online Media Services Pvt Ltd
Tags:
Journalism of Courage
A week after Google introduced its AI Overviews feature for Search, many users seem to have flagged issues related to it. The feature offers an AI-generated overview of topics or search terms a user looks for.
The feature which is currently only accessible to users in the US, has many in a fix as several users have reported that the AI-generated overviews were bizarre and unrelated to their searches, and some cases also included instances of misinformation.
Here are four such examples that indicate Google AI Overviews is far from perfect.
Someone wanted help with knowing how he can ensure cheese sticks to pizza. He made a Google search using the keywords ‘cheese not sticking to pizza’. Google AI Overview suggested adding non-toxic glue to the sauce to ensure the cheese sticks. The source for this tip is an 11-year-old post on Reddit.
When a user keyed in a simple query, “Is my employer required to give me a break during 8h shift,” AI Overview straight up said “No”, quoting The Fair Labor Standards Act (FSLA).
When a user searched about the health benefits of tobacco, AI Overview showed, “Tobacco contains nicotine, which can cause some short-term benefits such as increased alertness, euphoria, and relaxation. Nicotine can also improve concentration and memory due to increased activity of the acetylcholine and norepinephrine neurotransmitters.” While this is not factually wrong, no adverse effects of tobacco were mentioned.
When a user searched which US President attended the University of Wisconsin-Madison (UW-Madison) and earned 59 degrees, it mentioned 13 US presidents, which is absolutely wrong, as some of these people just have the names of US presidents.
In a recent interview with The Verge, Google CEO Sundar Pichai said he strongly feels that integrating AI into search will bring more value to users. He also said through AI previews that give context, people engage more and branch off, driving growth over time.
At this moment, AI Overviews, which derive responses from social media posts and blogs, do seem to have issues in providing authentic information. It is best to verify what you see on AI Overviews.
