Google advances with More interest. Remember Presentation of the AI, which essentially sums up the content drawn from websitesAnd presents it at the top of the Google search page?
This functionality subject to errors is now expansion to the American market, powered by the New gemini 2.0 ai models. It no longer requires Google account connection and has opened up to users of all age groups. Although this is a risky decision in itself, Google offers treatment similar to the whole search page with a new AI mode.
Currently available as a laboratory experience, AI mode essentially transforms the traditional Google research experience with website links to a conversational dialogue, the way AI chatbots give you answers. It is a delicious convenience, but could prove considerably wrong, if the history of AI previews is something to pass.
What is the AI mode for Google research?
The overall idea is to give users all the information they need – drawn from indexed websites – and save the hassles of clicking on sources and reading the web pages to find the answers. You can request monitoring requests in a natural language format, instead of a clogged search by keyword, and even provide details that would otherwise require some additional monitoring requests.
“He uses a” REQUIVE FAN-OUT-OUT-OUT “technique, issuing multiple related research simultaneously between sub-themes and several data sources, then brings together these results to provide an easy-to-understand response,” explains the company.
Google, however, warns that he may not always do things with AI mode, even if his internal tests have provided encouraging results. In scenarios where IA mode is not confident about the summary answer, it will simply display a list of web search results, such as the traditional Google research experience.
In his current avatar, he can provide answers as a well -formatted text or table wall, but on the road, images and videos will also be included. AI mode is currently available only for Google One AI Premium subscribers and will take place as an opt-in experience.
It is a bad omen for anyone who depends on Google research, especially if we are talking about precision. Here is an example. I looked for if we live in 2025.
A rich history of risks
The idea behind AI mode for Google research is theoretically rooted in the convenience of users. However, the fundamental technological pile behind it is still faced with some problems that the entire AI landscape has not yet resolved. One of them is the hallucinations of AI, which is essentially an AI tool invented information and presents it with confidence as a fact.
Google AI’s overviews are the best example of these missteps, and errors continue to arise to date. Take for example this evidenceWho was shared a few hours ago on Reddit, in which the AI preview lied confidently on a right conduct rule in India.
This is a false declaration, and yet, at no time, the language of the AI preview text suggests that the user should verify this information. “It’s so incomplete and so buggy that I am surprised that it even exists,” explains another report detailing his pure inaccuracy.
AI glimps appear only as a nugget condensed with information, served at the top of the Google search page. Now imagine an entire page that is presented to users as a long presentation, with some source links interspersed by the wall of the text.
Google says that IA previews excel in “coding, advanced mathematics and multimodal requests”. However, not so long ago, he tried the facts and turned the story on his head, in particular with the type of request in natural language which are being excited for the IA mode.
When request That astronauts met cats on the moon, he agreed with confidence that it was true, adding that astronauts were even taking care of these lunar cats. Virginia Tech Digital Literacy Expert, Julia Feerrar, do That AI does not really know the answers to our questions, quoting an example where Google I saw mentioned with confidence Barack Obama as the first Muslim president.
https://t.co/w09ssjvokj pic.twitter.com/6alcbz6ejk
– SG-R01 (@heavenrend) May 22, 2024
The consequences of disinformation of AI could be disastrous, especially with regard to health and well-being queries. In an analysis of more than 30 million pages of search engines (SERP), Sertstat have found that health related research is the most popular category where IA previews appear.
It is the same tool as suggested A person should eat at least one rock per day, adding an eighth cup of pizza glue, drinking urine to distribute kidney stones, and mentioned That a baby elephant can integrate into a human palm in 2025.
It is not the evolution of the research that I am looking for
Despite Google’s claims on how AI models have evolved, the situation has not improved considerably. There is less than a day, Futurism IA glimps have spotted confidently by saying that MJ Lenderman won 14 Grammy Awards.
It even hurt the year, when I asked for something as simple as “is it 2025” in the Google research area. “No, this is not currently the year 2025. The current year is 2024,” said IA overview.
Going a little further, he explained how 2025 is a current year that begins on Wednesday, adding a bunch of information that is not linked to discuss everything, from national celebrations to UN declarations which have absolutely nothing to do with my request.
Now I am not entirely against AI. On the contrary, I largely use tools like Gemini’s deep researchAnd often count on the latest Gemini 2.0 Flash model for creative ideas when my brain cells do not pull the advanced capacity.
However, pushing an overhaul of IA subject to errors to a source of information as essential as Google research is a risky proposal. Digital Trends has contacted Google and will update this story once we have obtained an answer.