Google offers a new feature with Artificial Intelligence: is it reliable for asking questions about health?

Recently, in May, Google launched a new feature in the United States called AI Overviews, which works with artificial intelligence Generative (AI) .

This tool collects information collected from all over the Internet and provides answers to questions asked by users.

Although presented as a promising proposition, in its first weeks of operation Different types of errors and inaccuracies have been reported in the data who shares.

That This was also observed in queries related to the domain of health which could put patients at risk if they follow inaccurate or downright false information.

For example, from New York Times They did a test and asked the system: “How many stones should I eat?” .

Faced with this request, The answer was that people should eat at least one a day in order to obtain vitamins and minerals.

As we know, The above is false and may have serious consequences in the body.

The AI ​​in question took such words from a satirical page and which, as one might imagine, had no medical credibility.

This is why the director of artificial intelligence at UC San Diego Health, Kaandeep Singh, told the aforementioned media: “You shouldn’t trust everything you read” .

On the other hand, Google senior product manager Hema Budaraju told the Times that have taken steps to ensure that health-related searches include “additional protective barriers” .

However, I haven’t described them in more detail. they told the American newspaper.

Although the answers mention some well-known sources in this area, such as the Mayo Clinic, WebMD, WHO, and PubMed, The summary provided by the AI ​​can also use sites like Wikipedia, Reddit, blogs or even commercial sites .

This without detailing which specific data comes from which sources .

An experiment they did in the Times was see the following: “Is it healthy to eat chocolate?” » .

Even if the response contained truthful information, it incorporated a text from a company dedicated to the sale of these products .

Likewise, when they asked the question another way: “Is chocolate healthy for you?” » – came across claims appearing on a site selling “home gut intelligence tests” and an app of food.

What awakens antibodies against this function with AI is that a block of text is presented that combines data from multiple sources, which could cause confusion and make it more difficult to assess veracity of these quickly.

Specialists in the field of health explained that in addition to this, does not distinguish between strong scientific evidence and weaker study results in addition to not including the necessary warnings for each case, if applicable.

Furthermore, Doubts remain as to whether it includes the most recent advances (and which have been validated by the scientific community) or whether it continues to show results that have already been refuted. .

For the director of the Stanford Health Communication Initiative, Seema Yasmin, This point is essential, because “science is not a set of static facts”. .

A Google spokesperson told the Times that Summaries provided by AI Overviews will match information provided by initial search results. .

With, stressed that its goal is to help users get an idea, but that the system is not designed to replace content delivered by specialized sites.

Along the same lines, a representative from the Cleveland Clinic pointed out that searching for health information on the Internet you must “research directly from known and reliable sources” .

It is good to remember that It is always advisable to consult a healthcare professional if you have any questions about your health or if you are experiencing symptoms of a possible illness. .