Google's new AI Overviews feature is giving many of its users bizarre and hilariously incorrect search results.
The company is scrambling to fix the issue after receiving tremendous backlash from users who flocked to social media to poke fun at the results they received.
AI Overviews offered inaccurate remedies for treating rattlesnake injuries, suggested applying glue to pizza when someone searched "cheese not sticking to pizza," and wrongly stated that former President Barack Obama is a Muslim.
A rock a day keeps the doctor away
byu/Darth_Vaper883 ingoogle
One user searched "What is in Google’s AI dataset?" and was met with a shocking response that the AI model was trained on child sexual abuse material.
Another instance saw AI Overviews giving the wrong answer for simple mathematical conversions.
That’s really bad.
— Lily Ray 😏 (@lilyraynyc) May 27, 2024
Now Google doesn’t even know that cm = centimeter?
And the calculator appears automatically with a wrong answer? Yikes.
(I was able to replicate) https://t.co/DgOcfCp2QF
Google has since disabled AI Overviews for specific searches due to these unexpected challenges.
The tech giant reassured critics that it is taking swift action to solve these problems while still maintaining that AI Overviews generally provide high-quality information.
Growing Challenges, Concerns About AI
Google’s project has been under testing for a year, as part of the Search Generative Experience launched in beta in May 2023.
In a statement to The Verge, CEO Sundar Pichai said that the feature handled over a billion queries during this period, achieving multiple breakthroughs in hardware and engineering.
Google announced during its I/O developer conference on May 15 that it will start rolling out AI Overviews in the U.S.
AI expert Gary Marcus, an emeritus professor of neural science at New York University, spoke about the challenges faced by AI companies.
He pointed out that while achieving 80% accuracy in an AI model is relatively straightforward, reaching the final 20% is exceedingly difficult.
He, along with Meta Chief AI Scientist Yann LeCun, believes that the current large language models like Google’s Gemini and OpenAI’s GPT-4 will not be enough to achieve artificial general intelligence (AGI).
General intelligence, artificial or natural, does not exist.
— Yann LeCun (@ylecun) May 26, 2024
Cats, dogs, humans and all animals have specialized intelligence.
They have different collections of skills and an ability to acquire new ones quickly.
Much of animal and human intelligence is acquired through… https://t.co/RqK8zxaf7E
Concern is also growing over the wider ramifications of AI-generated content.
Experts say that 90% of content on the internet will be AI-generated by 2026, emphasizing the need for quality control and content verification in AI systems like Google's.
As Google covers up this latest mishap, experts and consumers alike are beginning to doubt the viability of AI for public consumption in its current state.
READ NEXT: Scarlett Johansson Slams OpenAI for Voice Usage
Editing by Katherine 'Makkie' Maclang