- Tips for Spending Holiday Time With Family Members Who Live with Dementia
- Tainted Cucumbers Now Linked to 100 Salmonella Cases in 23 States
- Check Your Pantry, Lay’s Classic Potato Chips Recalled Due to Milk Allergy Risk
- Norovirus Sickens Hundreds on Three Cruise Ships: CDC
- Not Just Blabber: What Baby’s First Vocalizations and Coos Can Tell Us
- What’s the Link Between Memory Problems and Sexism?
- Supreme Court to Decide on South Carolina’s Bid to Cut Funding for Planned Parenthood
- Antibiotics Do Not Increase Risks for Cognitive Decline, Dementia in Older Adults, New Data Says
- A New Way to Treat Sjögren’s Disease? Researchers Are Hopeful
- Some Abortion Pill Users Surprised By Pain, Study Says
Doctors Still Beat AI in Offering Accurate Medical Advice: Study
It might be too soon to rely solely on machine learning for health advice, a new study finds.
After pitting the latest AI against actual human physicians, the doctors easily won, reports a team led by Dr. Andrei Brateanu, of the Cleveland Clinic Foundation.
For now anyway, “AI tools should not be seen as substitutes for medical professionals,” he said. “Instead, they can be considered as additional resources that, when combined with human expertise, can enhance the overall quality of information provided.”
As the researchers noted, consumers can already harness the power of web searches or AI programs like ChatGPT-4 or Google Bard whenever they have medical questions.
But how accurate are the answers?
To find out, Brateanu’s team posed 56 common medical questions to either ChatGPT-4 and Bard. They then had two physicians review the answers for accuracy (a third doctor was brought in to settle things when the two doctors disagreed).
Of the two AI programs, Bard seemed to outperform ChatGPT, although both had relatively low levels of accuracy, the study found.
For example, the physicians found 53.6% of Bard’s answers accurate, 17.8% inaccurate and 28.6% partially accurate.
ChatGPT performed worse: Just 28.6% of answers were accurate, 28.6% inaccurate, and 42.8% were partially accurate but incomplete, the researchers said.
They reported the findings recently in the American Journal of Preventive Medicine.
The bottom line, according to Brateanu: AI programs “can offer medical information that may look very accurate and convincing, when in fact it may be occasionally inaccurate.”
“The fact that both models produced responses with inaccuracies or omitted crucial information highlights the ongoing challenge of developing AI tools that can provide dependable medical advice,” he added in a journal news release.
His advice? Head to AI for medical information, but always run it by your doctor afterwards.
“As we incorporate AI technology into healthcare, it’s crucial to ensure that the essence of healthcare continues to be fundamentally human,” Brateanu said.
More information
Find out more about the role of AI in health care at the Pew Research Center.
SOURCE: American Journal of Preventive Medicine, news release, April 1, 2024
Source: HealthDay
Copyright © 2024 HealthDay. All rights reserved.