AI in Medical: Recently a research has come to light. In this, AI performed better than some doctors in diagnosing patients’ diseases and planning their treatment. Even in emergency rooms, AI saw many cases and in most cases performed better than doctors.
AI is working better than doctors
In a new research published in the journal ‘Science’, it was told that 76 clinical cases from the emergency room of Israel Deaconness Medical Center were given to OpenAI’s o1 model and two expert doctors. Researchers found that o1 performed equally or much better than human experts in a variety of tasks. When the least information was available, o1 identified the disease at or very close to correct in 67.1% of cases. Whereas the accuracy of both the doctors was 55.3% and 50%.
As the doctor got more information about the patient. The model’s accuracy increased to 72.4%, while the two doctors’ accuracy was 61.8% and 52.6%. The accuracy of o1 reached 81.6% when patients were admitted to the medical floor or ICU. This time too it left both the doctors behind.
The work can be risky
The study also found that when AI was asked to plan treatment such as prescribing antibiotics or making end-of-life decisions, it was ahead of doctors. Across five case studies, AI achieved an average score of 89%, which was significantly higher than doctors’ scores. Doctors scored about 34% using traditional methods and 41% using GPT-4.
The researchers said that using AI to help make clinical decisions is sometimes considered a high-risk task, but greater use of these tools could help reduce diagnostic errors, delays and lack of access to treatment.
There is a need for such research
However, the researchers also warned that clinical medicine is full of textual input. Such as the level of physical discomfort of patients or information about medical imaging. He suggests that this means that in the future there is a need for researchers who can assess how humans and machines can work together effectively.