Could an AI Diagnosis Be Better Than a Human’s?

From banking, to legislation, and even to creating art, artificial intelligence, or AI, is exponentially developing and is becoming commonplace in many of our lives. Starting in the mid 1900s, the idea of technology thinking for itself sprung into the mainstream, with the Tin Man from the Wizard of Oz considered as the first time the general public encountered the concept of artificial intelligence. Since then, a fear has grown that robots will become more intelligent than humans and eventually replace them. With self-checkouts at grocery stores and cars learning to drive on their own, this fear of technology replacing humans is beginning to feel more like reality.

With artificial intelligence technology becoming more advanced year after year, this question of if AI could take over human jobs has begun to creep into more knowledge-based industries, like healthcare. One interesting study tested the performance of the AI language model, ChatGPT, on the United States Medical Licensing Exam (USMILE). 

Amazingly, the model passed all three sections of the exam, although it did so with slim margins. However, what is most significant about this study is that this was a new language model, launched only two months before completing the exam. Furthermore, it had not been trained on any biomedical data. Essentially, it passed without studying. 

This shows enormous potential for AI-assisted education and perhaps even the ability to provide patients an accurate AI diagnosis. However, it also begs the question, could AI become better at diagnosing than a human healthcare professional? With this in mind, we asked the Figure 1 community for their thoughts on AI in healthcare.

Here is what more than 700 healthcare professionals had to say.

Would You Use AI?

With the revelation that AI can perform exceptionally well on medical exams, we asked the Figure 1 community if they would consider using artificial intelligence in their clinical decision making. The results are quite positive with 55% of respondents saying yes, they would use AI. The remaining 45% of responses were split quite evenly with 23% responding no, and 22% responding that they were unsure. 
It is important to note that many respondents see AI as a new tool in the toolbox rather than a substitute for decision making. One respondent stated, “I see AI as another tool in medical decisions. You input your information about the patient and the system gives you some suggestions as to what you are looking for and a course of treatment.”

Should AI be Implemented?

It is one thing to ponder the abilities of a new tool and think about how much potential it could have in the future. It’s completely different to think about how a tool would pragmatically be used. So, we asked our respondents if they thought AI software is ready to be implemented into clinical practice. In contrast to our first question, where many respondents were open to the idea of using AI in clinical decision making, the majority of respondents believe that AI is not ready for use just yet. A majority of 45% said no, AI is not ready for clinical decision making, and 26% remained unsure, leaving only 29% believing AI could be used in practice today.

AI vs. Humans

The question everyone asks when it comes to artificial intelligence is, will AI be better than humans? So we asked the Figure 1 community, when it comes to clinical decision making, do you think an AI diagnosis will be better than a human’s? Surprisingly, this question had the most mixed results of all. Thirty-eight percent of respondents said yes, 43% said no, and the remaining 19% were unsure. 

Overall, it seems respondents lean slightly in the direction that technology will never be able to take on the full scope of what it means to be a healthcare professional. As one registered nurse put it, “AI is an excellent adjunct to the current tools we have to diagnose patients. However, it will not replace the human factor in assessing and diagnosing, as one can sometimes notice small nuances that would bypass the AI.”

Published March 13, 2023

Join the Conversation

Sign up for Figure 1 and be part of a global community of healthcare professionals gaining medical knowledge, securely sharing real patient cases, and improving outcomes.