IT doctor Isaac Kohane has been testing GPT-4 with two colleagues. It had one main goal: to find out how OpenAI’s latest AI model would perform in a medical setting.
I was shocked to say: better than many doctors I know
Cohen wrote in the next edition The artificial intelligence revolution in medicine (The Artificial Intelligence Revolution in Medicine), co-authored by journalist Carrie Goldberg and Microsoft research vice president Peter Lee.
the inside According to the In the book, Isaac Cohen claims it was released to subscribers in March 2023 GPT-4 More than 90 percent of the questions on the US medical licensing exam are answered correctly.
GPT-4 is also a great translator and can even translate a complex scientific term into something even sixth graders can easily understand.
He can imitate doctors with great success
As the authors explain through live examples, GPT-4 can provide clinicians with helpful tips on how to talk to patients empathically about their disease, and they can read long reports or studies and summarize them at a glance.
In addition, he can provide explanations for certain problems in a way that requires some ability reminiscent of human intelligence.
When GPT-4 was asked by the authors of the book if he was truly capable of causal reasoning, he said “All his intelligence is limited to patterns in the data, with no real understanding or intention involved.” Even with these limitations, Cohen writes, GPT-4 can mimic the way doctors diagnose disease with impressive success.
Can doctors be replaced by robots?
As we wrote, although doctors will not be replaced by robots due to the rise of artificial intelligence, there are quite a few who may be concerned. AI expert Dr. György Tilesch previously spoke to Index about the risks of the technology, its regulatory background, and even its galactic side.
The expert said: The era of the second fatwa will come. Although AI can also be used in the medical profession, for example it can diagnose, at the same time it is important to use this only as a second opinion, so we rely on the doctor’s findings and compare the robot’s insights to that, and if there are differences between the two, we explain We see how correct the program’s conclusions are.
At this point, I’d like to mention that last year, Google announced its own Doctors’ Lounge, which provided faster and more accurate medical answers to patients than a doctor in a white coat. Not to mention that the robot does not go to lunch, does not get tired, and does not ask for time off
he added.
