DREAM / The Prime Minister resigns and wakes up under the Parisian bridge!

 Kliko këtu për të tjera
Voxpress:

Can you trust ChatGPT's medical advice?

2025-10-03 14:15:44Lifestyle SHKRUAR NGA REDAKSIA VOX

Can you trust ChatGPT's medical advice?

“What is lupus?” “How long does the flu last?” “How do you treat hemorrhoids?” These are some of the most common health questions people are asking ChatGPT.

The popularity of large language models (LLMs) like ChatGPT for delivering personalized health advice is growing. One in ten Australians now use the platform to ask medical questions, according to a survey of nearly 2,000 Australians conducted in mid-2024. The study, published on Tuesday, found that almost two in three people (61%) who use ChatGPT for medical advice ask questions that typically require clinical advice.

"Artificial intelligence tools are popular because they can provide quick answers to any question. [However], as with all these tools, there is always the risk that they will give the wrong answer," said study author Julie Ayre from the University of Sydney.

With so many people using artificial intelligence models to ask about their health status, can they be trusted?

Researchers are building a scientific consensus about the (un)reliability of medical advice from LLMs, however the findings quickly become outdated as new models with better algorithms are published and updated.

A study in 2024 challenged ChatGPT3.5 with 150 medical cases, including patient history, symptoms, and hospital test data, and asked the artificial intelligence to make diagnoses and a treatment plan.

The results weren’t great. ChatGPT only provided the correct diagnosis and treatment plan 49% of the time, making it an unreliable tool. The authors concluded that ChatGPT “does not necessarily provide factual accuracy, despite the large amount of information on which it was trained.”

Another study concluded that ChatGPT “did not reliably provide appropriate and personalized medical advice,” but could provide relevant information for medical questions.

When researchers assessed the quality of medical information on ChatGPT in a 2023 study, they asked ChatGPT3.5 “why should you treat jaundice caused by gallstone disease?” The study responded that relieving jaundice improves the patient’s appearance, and this improves self-esteem.

"This is not really the clinical rationale," said Sebastian Staubli, a surgeon at the Royal Free London NHS Foundation Trust, UK, who led the study.

The newer ChatGPT4.0 provides better answers to the question, emphasizing the need to prevent organ damage and disease progression. The problem with ChatGPT is that, while its medical advice is not completely incorrect, it is not completely accurate either.

The quality of the information on which an artificial intelligence model is trained determines the quality of its medical advice. The problem is that no one knows exactly what information specific models are trained on.


Në përputhje me standarded JTI

Video

SHIKO KANALIN NE YOUTUBE
×