A study conducted by scientists at Long Island University (USA) and published on the university’s official website identified potential shortcomings in the public version of the chatbot ChatGPT when providing accurate medical information.

As part of the experiment, scientists asked the chatbot 45 questions related to medical aspects. The results showed that the bot refused to directly answer 11 questions and gave incorrect answers to 22. For each question, the researchers asked for sources, but ChatGPT provided links in only 8 cases, and all of them were invalid.

Particular attention has been drawn to the issue of a possible interaction between the antiviral drug for COVID-19, paxlovid, and the blood pressure drug verapamil. The bot claimed there was no interaction, when in fact using them together could lead to a dangerously excessive drop in blood pressure.

The scientists noted that they used the free version of the chatbot, accessing data downloaded and processed no later than September 2021. They suggested that the paid version of the tool was likely to be more accurate in providing medical information.

Previously reportedthat scientists have discovered the effectiveness of physical activity for older people.