ChatGPT passed the Chinese Medical Practising Medical license exam with high scores. Is AI going to treat you in the future?

Chao News Client Reporter Zhang Miao, Zhu Ping, Correspondent Fang Xu, Zhu JunjunCan you imagine being treated by an artificial intelligence robot in the future?It is not a rare scene in science fiction movies. It seems that there are more and more possibilities to become a reality

Chao News Client Reporter Zhang Miao, Zhu Ping, Correspondent Fang Xu, Zhu Junjun

Can you imagine being treated by an artificial intelligence robot in the future?

It is not a rare scene in science fiction movies. It seems that there are more and more possibilities to become a reality. Last month, the team of Wu Chenghan, Neurosurgery Department of the Second Affiliated Hospital of Second Affiliated Hospital of Zhejiang University School of Medicine, published a paper in the journal Digital Medicine, and they passed the Chinese Medical license examination with ChatGPT.

After ChatGPT passed the American Medical Practitioner Examination in February of this year, this generative AI, who was only born for more than six months, passed the highly valuable qualification certificate national examination in Chinese.

So the question arises, will generative AI, which has demonstrated its expertise in various fields, replace doctors?

Image source: Visual China

Pass the "random exam" with high scores

First, let's see how ChatGPT passed the Chinese Medical license examination.

The examination of Medical license is divided into two parts: practical skills examination and written examination.

In the practical skills exam, Wu Chenghan's team sent 10 medical history collection questions to ChatGPT and invited two medical students to answer together. Finally, the three unsigned answers were handed over to three doctors for evaluation and scoring.

This part has a maximum score of 15 points, and ChatGPT has received 11 points, which is not high, but it has already passed. "Wu Chenghan said, and the team found that ChatGPT can cover all clinical consultation content in detail and focus on the key points of the consultation.

Next is the written exam, where the team tested the ChatGPT using the Chinese Medical Practitioner Qualification Examination over the past three years and found that it could correctly answer about 72% of the questions. This score is higher than 80% for Chinese exam oriented medical students.

Wu Chenghan was surprised at such a result. The reason why he was surprised was not that AI passed the Medical license exam, but that ChatGPT, which uses English as the main training language, was not a professional model trained for any exam, and could also pass the Chinese exam. From the final results, it can be seen that the scoring rate of ChatGPT is similar in both knowledge and logic types, in other words, when it is targeted for training, its score will be higher.

Image source: Visual China

What does it mean to pass the practicing physician exams in China and the United States?

At that time, ChatGPT passed the three most challenging professional exams in the United States, including the American Medical Practicing Examination (USMLE), the Lawyer Qualification Certificate exam, and the Wharton Business School MBA exam. "Professor Zhang Jianmin, director of the Brain Science Center at Zhejiang University Second College and director of the Institute of Brain Medicine at Zhejiang University, believes that" behind the high score of ChatGPT, there are various complex factors

The American Medical Practitioner Qualification Examination is a standardized examination that typically requires medical students to spend 300 to 400 hours of professional study time completing the first part, not to mention two additional parts that involve clinical diagnosis, medical management, and internship practice, Without participating in pre exam review classes and intensive classes, ChatGPT easily achieved a score rate of 52.4% to 75%, which is better than most students and indeed has strong abilities. However, when seeing the results, Zhang Jianmin emphasized that "factors such as test question processing methods cannot be completely disregarded. For example, in ChatGPT's scores, vague answers are removed and standardized answers are emphasized

Zhang Jianmin's view coincides with an article published in Public Library of Science Digital health earlier this year. In this article titled "ChatGPT Draws Attention to the Shortcomings of Medical Education through the American Medical Practitioner Qualification Examination", it is proposed that the success of ChatGPT reflects on the one hand, the excessive emphasis on mechanical memory in physician exams, rote memorization of disease mechanisms, and "insufficient evaluation of the skills required for modern medical practice"; On the other hand, it reflects the rigidity of medical teaching methods, which can easily make students mistakenly think that medical problems are either right or wrong.

In clinical diagnosis and treatment of diseases, doctors need to follow diagnostic and treatment standards, but on the other hand, they must also be tailored to the specific situation of each individual, which is the principle of personalization. Because each patient's social background, family background, traditional habits, economic conditions, and personal preferences are different, facing the same disease, doctors also need to consider many practical factors in order to make personalized treatment decisions with humanistic care. Sometimes patients reveal these information only through a tiny expression, or even a sigh in the conversation. "These small actions may affect the doctor's treatment decisions, but obviously AI doctors cannot perceive them."

Doctors' judgment of a disease still needs to be analyzed from a dynamic perspective to see the evolution process of the disease. In Zhang Jianmin's view, the current diagnosis provided by artificial intelligence is more based on basic conditions and rule-based and regular data to determine what diagnosis this is and how to treat it. It is more like learning and memory within a three-dimensional range, and our doctors often need to incorporate the time dimension into the diagnosis, so that we can see the problem more comprehensively

Although more and more studies have shown that ChatGPT has an increasing probability of being reasonable in medical practice, there is still a possibility of an unreasonable answer, which may have adverse consequences for serious and life related medical fields.

So, while marveling at the amazing learning ability of ChatGPT, Zhang Jianmin believes that there is still a long way to go before artificial intelligence can be fully utilized in hospitals.

Image source: Visual China

Will doctors lose their jobs as a result?

Artificial intelligence cannot replace doctors, but it can definitely help doctors better, especially improve local medical standards. "Wu Chenghan's answer is very firm.

The reason is simple. In the medical process, the participation of "people" is needed as a closed-loop.

Wu Chenghan mentioned, "For example, if experienced department heads participate in the practicing physician exam, they may not necessarily get full marks, but their error rate in their daily medical work may not be less than one in ten thousand." In his view, exam scores and medical technology level cannot be equated, and the gap between them is because more "people" in the medical team are at work.

So, how will AI, which has easily passed the Medical license, help doctors and patients?

Just this week, the Ruijin Hospital affiliated to Shanghai Jiaotong University School of Medicine demonstrated the latest achievements in the construction of the "future hospital", including AI health consulting enabled by the big model, "Ruijin on the cloud" intelligent services, and the intelligent operation platform of full process control, which made people see the more powerful functions of "artificial intelligence+medical treatment" and the various possibilities of future medical treatment.

What we need to utilize is the intelligence of artificial intelligence, but it cannot solve multidimensional problems. It is destined that artificial intelligence will not have high intelligence like humans. It is not omnipotent, "Zhang Jianmin believes, The medical field is just one of many fields after the emergence of AI, but medical treatment has its particularity and is face-to-face with people. Therefore, artificial intelligence needs to further pay attention to its safety, effectiveness, and ethical regulatory issues, and require corresponding regulatory guidance, laws and regulations, and corresponding regulatory measures

In the future, although ChatGPT may not necessarily replace doctors in diagnosing and treating patients, future healthcare will definitely be human-machine collaborative.

This has also received a response from Google, one of the earliest companies to study AI doctors. On July 12th, researchers from Google and DeepMind published a study in the journal Nature, stating that AI doctors built on large language models are in many ways comparable to human doctors, but this technology will not pose a threat to the work of general practitioners and may be used more for medical helplines in the future.

Image source: Visual China

What are our expectations for artificial intelligence healthcare?

Have you ever had such an experience? Hang up the expert account, queue for 1 hour, and finally communicate with the expert for less than 10 minutes... "We hope that artificial intelligence can play a role in interacting with patients and collecting medical history," said Wu Chenghan. In April of this year, a research team from the UK used ChatGPT to write clinical reports for patients. They found that ChatGPT can provide relatively accurate disease science and health guidance in a short period of time, and its writing is more compassionate and patient compared to most reports written by clinical doctors.

I hope that artificial intelligence can become a bridge between patients and doctors, which is also one of the reasons why Wu Chenghan's team started this research.

In the central hospital of Zeya Town, Ouhai District, Wenzhou, you can see the scene that Lu Pengfei, a young general practitioner, is wearing an AR glasses, and from the loudspeaker, another more mature male voice is emitted.

Lu Pengfei is connecting with experts from the superior hospital through AR glasses to see patients.

Our superior hospital experts can clearly see the patient's condition and make timely diagnosis through such AR equipment, "said Lu Pengfei.

In this technologically rich scenario, there is no "artificial intelligence" at work, but in Lu Pengfei's view, the development of artificial intelligence can make remote consultations with such experts more common.

We do not lack patients, but sometimes there is a lack of diagnosis and medication for difficult cases. Connecting experts through AR glasses is a way, and artificial intelligence is also a way to judge difficult cases and provide medication advice

"At the first time of our emergency, artificial intelligence should not work, but it can play a great role in the follow-up treatment." Lei Guoxin, a doctor in the Emergency physician of Songyang County People's Hospital, told Chao News, As a grassroots doctor, we always have to face new medical situations, but there is a lack of communication with experts and updated medical knowledge. For example, when a new specific drug appears, which components of the drug are in conflict or incompatible? Solving this problem is both experience and guidance, and artificial intelligence can provide us with help in this regard

Although ChatGPT has not yet been able to serve as a drug index, Lei Guoxin has used it as a tool for answering questions. He said, "In order to pass the intermediate professional title exam, we will download the relevant question bank app and buy a member, but ChatGPT can already set questions for us and provide answers

The auxiliary role of basic level medical care to artificial intelligence is eagerly anticipated, and in the sophisticated research of human medical care, artificial intelligence is also making efforts, especially in the three fields of Medical imaging recognition, disease diagnosis and pharmacy.

In terms of Medical imaging recognition, AI is mainly to help doctors pre label the parts of the image that may have lesions. This saves a lot of time, and at the same time, the advantages of artificial intelligence in terms of accuracy and efficiency are evident in some difficult to detect diseases.

In the United States, a dental care company called VideaHealth has pioneered the application of artificial intelligence in the dental field. Currently, 90% of dental clinics in the United States can use this platform to assist in diagnosis.

In addition, artificial intelligence also has extremely impressive performance in disease diagnosis.

A study in the internationally renowned scientific journal PLOSONE pointed out that deep learning models are more accurate in the diagnosis of Alzheimer's disease than untrained artificial intelligence models, and this technology can exclude some factors that interfere with diagnosis, such as age.

Researchers at Massachusetts General Hospital trained a deep learning model using tens of thousands of human brain scan images, some of which were the brains of patients with Alzheimer's disease, while others were not. The study then compared the model with real clinical diagnostic data.

In the field of pharmaceuticals, the role of artificial intelligence is focused on the development of new drugs. With powerful data integration and analysis capabilities, artificial intelligence can help scientists identify new drug targets, design new drugs, predict drug efficacy and toxicity, thereby shortening drug discovery and development time and saving costs.

Please indicate the source for reprinting

Disclaimer: The content of this article is sourced from the internet. The copyright of the text, images, and other materials belongs to the original author. The platform reprints the materials for the purpose of conveying more information. The content of the article is for reference and learning only, and should not be used for commercial purposes. If it infringes on your legitimate rights and interests, please contact us promptly and we will handle it as soon as possible! We respect copyright and are committed to protecting it. Thank you for sharing.(Email:[email protected])