Google’s latest AI feature, “AI Overview,” has recently faced intense scrutiny due to a series of high-profile errors that have raised major concerns about its reliability.
This feature, designed to enhance the search experience by providing succinct summaries, has generated numerous inaccuracies that range from humorous to dangerously misleading, provoking widespread backlash from users and experts alike.
Here are a few glimpses of the errors generated by AI overviews.
I saw someone else mention this one on another post
Ah yes , the snake very famous for being a mammal 🤣 pic.twitter.com/xo538FcqoK— Raven – 🍉 (@RavenTenebris13) May 20, 2024
The AI Overview has made several glaring mistakes, such as recommending glue as part of a pizza recipe and suggesting that ingesting rocks could provide nutritional benefits.
Post by @miasato.2View on Threads
One instance involved the AI advising users to mix non-toxic glue into pizza sauce to prevent the cheese from sliding off, which was traced back to a decade-old joke on Reddit.
In another case, it told users to consume rocks for vitamins and minerals, advice originating from a satirical article by The Onion. These errors not only amused the public but also raised serious concerns about the reliability of AI-driven search tools.
Source: NBC News
While humorous errors in Google’s AI search have captured attention, they also highlight deeper issues affecting the platform’s efficacy and user trust. For an in-depth discussion, read our analysis on the decline of Google search.
These incidents have undermined trust in Google’s search engine, which more than two billion people rely on for accurate information. The backlash highlights Google’s unique pressure to integrate AI into its services safely.
1/ In recent weeks, Google has introduced “AI Overviews” in their search results. This move seems to be a direct response to the rapid growth and adoption of ChatGPT by OpenAI hurting Google search traffic. Let’s explore why this is happening. #Google #AI #ChatGPT pic.twitter.com/M9PHj0kZyr
— jimboot (@jimboot) May 27, 2024
Despite these errors, Google has emphasized that the vast majority of responses generated by AI Overview are of high quality and that the problematic answers are isolated examples often stemming from uncommon queries. However, NBC news was able to reproduce several incorrect responses, further questioning the reliability of the tool.
(Source: NBC News)
Here’s another error!
(Source: NBC News)
Similar issues have marred Google’s history with AI features. For example, the initial rollout of the Bard Chatbot provided incorrect information about outer space, leading to an enormous drop in the company’s market value. Bard’s successor, Gemini, also faced criticism for its inability to generate accurate images of historical figures and its refusal to depict white individuals in certain contexts.
I think AI Overviews will be like Google Plus. pic.twitter.com/5remHSYbEZ
— James Norquay (@connections8) May 25, 2024
Tech industry insiders have criticized Google for these repeated blunders, arguing that the company is rushing its AI developments to keep pace with competitors like Microsoft and OpenAI.
Explore a detailed overview of some of the worst responses by Google’s AI Oveview.
Financial analysts suggest that Google has little choice but to move quickly, even if it means dealing with growing pains along the way.
Thomas Monteiro, a Google analyst at Investing.com, says, “Companies need to move really fast, even if that includes skipping a few steps along the way. The user experience will just have to catch up.”
Google spokesperson Lara Levin stated that the company is working to refine its AI systems and address these issues. She noted that many of the erroneous responses were due to doctored examples or queries not representative of typical user searches.
Levin reiterated that the company remains committed to improving the accuracy and reliability of AI Overview.
The errors have also led to broader concerns about the future of AI in search. The ability of AI to distinguish between factual information and satire remains a significant challenge, as highlighted by the use of content from The Onion as factual responses.
This is what users have to say about it.
Google LLM will never be accurate because it’s trained on reddit data + countless websites with non verified claims. On top of that, google LLM doesn’t know if something is a misinformation, joke, sarcasm or a troll. It takes everything at face value. Recipe for a disaster.
— Boris Razmiki (@slickd0g) May 24, 2024
Social media users have shared examples of the AI providing dangerously incorrect advice, such as mixing chlorine bleach and white vinegar for cleaning, which can create harmful chlorine gas.
That’s alarming, funny thing is, you still have people believing these wrong information.
— BaseGPT (@base_gpt) May 26, 2024
Former Google AI ethics researcher Margaret Mitchell has pointed out the potential dangers of AI-generated misinformation, emphasizing the need for more rigorous testing and oversight.
“This is about pointing out clearly foreseeable harms before—e.g., a child dies from this mess. This isn’t about Google, it’s about the foreseeable effect of AI on society,” Mitchell said.
The broader tech community remains divided on the future of AI in search, with some optimistic about its potential and others wary of its current shortcomings.
Concerns about factual inaccuracies are just the beginning; there’s also the broader issue of what AI-generated content means for those who make their business publish information on the web.
Gartner forecasts a 25% decline in search engine volume by 2026, which could greatly impact web publishers.
Comment
byu/Randomlynumbered from discussion
intechnology
Google CEO Sundar Pichai argues that AI summaries are better for web publishers in the long run, stating that “If you put content and links within AI Overviews, they get higher clickthrough rates than if you put it outside of AI Overviews.”
Despite these assurances, the errors continue gaining visibility and hype, with some social media users creating doctored responses highlighting the AI’s shortcomings.
Comment
byu/Randomlynumbered from discussion
intechnology
As Google continues to refine its AI tools, the tech industry and the public will closely watch how these developments unfold.
Integrating AI into search is a big step toward the future of information retrieval, but maintaining users’ trust and upholding Google’s reputation for informational integrity will require addressing these critical errors.
For more news and trends, visit AI News on our website.
