US EditionPublished • loading... • Updated
AI No Better than Other Methods for Patients Seeking Medical Advice ...
Nearly 1,300 U.K. participants identified health problems correctly only one third of the time using AI chatbots, matching results from traditional internet searches, study found.
- On Monday, the study published in Nature Medicine found AI chatbots do not provide better health advice than traditional methods and matched search engine results.
- British-Led researchers set out to test how well nearly 1,300 U.K.-based participants identified health issues using chatbots in 10 different scenarios, noting real users often withheld relevant information.
- On Monday, participants were randomly assigned to OpenAI’s GPT-4o, Meta’s Llama 3 or Command R+, or a control group using internet search engines, with diagnosis success around a third and correct action at around 45 percent.
- Rebecca Payne, Oxford University, said `Patients need to be aware that asking a large language model about their symptoms can be dangerous, giving wrong diagnoses and failing to recognise when urgent help is needed`, and David Shaw, Maastricht University, advised trusting reliable sources like the NHS.
- One out of every six U.S. adults consult AI chatbots for health, while researchers pointed to a communication breakdown between benchmarks and real users.
Insights by Ground AI
32 Articles
32 Articles
The next time you want to consult “Dr. ChatGPT,” you might want to think again. Although they can now excel in most medical licensing exams, AI chatbots don’t… Research: AI chatbots give wrong health advice - ΘΑΦΤΕΜΟΠΟΡΙΚΙ
Coverage Details
Total News Sources32
Leaning Left5Leaning Right4Center6Last UpdatedBias Distribution40% Center
Bias Distribution
- 40% of the sources are Center
40% Center
L 33%
C 40%
R 27%
Factuality
To view factuality data please Upgrade to Premium
