ChatGPT: Innovation or gimmick? Talk about its impact on the field of autoimmunity | Topic

Yimaitong edited and sorted, please do not reprint without authorization.

Since the vigorous development of science and technology, its achievements and innovations have gradually changed people's lives and cognition. The latest example is ChatGPT, a new artificial intelligence (AI) product released to the public on November 30, 2022. Artificial intelligence (AI) was first proposed by mathematician Alan Matheson Turing in 1950 to describe the concept of using computers for critical thinking and other human-like intellectual abilities. AI was first applied to medicine in 1976, when the possibility of using computer analysis to diagnose acute abdominal pain was explored.

Over the past two decades, modern medicine has used AI in many jobs, such as supporting clinicians in diagnosing, predicting the prognosis of certain diseases, and even formulating treatment options. The field of autoimmunity is a fast-growing and wide-ranging discipline, and its role in COVID-19 has made it one of the most studied topics over the past few years. Recently, the journal Autoimmun Rev. (impact factor 17.390) published a review "ChatGPT and Autoimmunity - A New Weapon on the Knowledge Battlefield", which for the first time discussed the role of ChatGPT in autoimmune and other medical fields in an official paper, illustrating the latest applications and limitations of this technology.

ChatGPT's roots in medicine

Artificial intelligence is new, but it is developing rapidly, mainly because of its powerful computing power. In fact, chatbots or large language models (which understand and respond to user input in a human-like way, such as ChatGPT) have been around for some time and are not recent innovations in AI. ChatGPT, an AI-based language chatbot that belongs to the natural language processing model that provides complex and "human" answers based on a given task, has received a lot of attention since its release.

On December 19, 2022, Victor Tseng of Acebo Health in California and Tiffany H. Kung of Massachusetts General Hospital published a paper announcing their assessment of ChatGPT's performance in the U.S. medical licensing exam: ChatGPT met or approached the passing threshold in all three stages of the U.S. medical licensure exam without any professional training or reinforcement. In addition, ChatGPT has shown a high degree of consistency and insight. This has undoubtedly re-stimulated the debate on whether artificial intelligence will eventually replace doctors, thus completely subverting the discussion of contemporary clinical medicine. In this regard, some scholars believe that large-scale language processing models have great potential to assist medical education and even clinical decision-making.

Potential applications of ChatGPT in medical fields such as autoimmunity

The potential uses of ChatGPT are wide-ranging, from helping academics with academic writing to writing code for software engineers. Since ChatGPT is a large language model, it can provide straightforward answers instead of giving hundreds of sources and/or links like Google or other search engines. Therefore, it can be used as an aid for researchers when writing papers. Objectively, ChatGPT may help clinicians relieve themselves of the pain of "climbing the grid" to write papers. But this caused great controversy, leading some scientific journals to ban ChatGPT as an author. For example, Nature prohibits large language model tools (LLM) including ChatGPT from attributing paper authors out of prudence, and if they are actually used, they should be clearly stated in the paper method or acknowledgement section. At the same time, ChatGPT can also help researchers in multiple aspects such as study design, data analysis, and recommended methods, and it can even screen out key strengths or weaknesses in certain studies to help researchers improve efficiency.

From a clinical autoimmune and rheumatological perspective, the potential use of ChatGPT is to provide and summarize relevant guidelines and treatment options, including the potential benefits of treatment, adverse effects, and drug interactions. Autoimmune diseases are often difficult to diagnose, and ChatGPT can help clinicians recommend certain diagnostic tests, such as laboratory or imaging tests, to help narrow down the differential diagnosis.

ChatGPT can play a role in managing electronic health records (EHRs), helping clinicians with tasks such as data entry, data extraction, and data analysis. In addition, it can educate patients about their condition and lifestyle, and even provide advice on nutrition and exercise programs. Unlike Google and other search engines, it is difficult for patients to "self-diagnose" through ChatGPT because it only provides patients with general disease information and advises them to consult a medical professional for evaluation and diagnosis.

Limitations of ChatGPT in the medical field

In parallel with the potential value, the current ChatGPT (with the fourth generation version as an example) also has its obvious limitations, such as incomplete information as of September 2021, lack of personalization, privacy concerns, insufficient clinical validation, and undiagnostics.

Throwing the question "Limitations of ChatGPT to the medical field" to ChatGPT, here's the answer:

• Data quality: ChatGPT mainly learns and generates text through data on the Internet, but these data are of varying quality and may have errors and biases, which may affect the accuracy and reliability of ChatGPT.
• Diagnosis and treatment options: ChatGPT can provide certain diagnosis and treatment recommendations, but for some diseases and symptoms, ChatGPT may not provide accurate advice, which may lead to health risks to patients.
• Ethical issues: In the field of medicine, it is very important to protect the privacy and rights of patients and patients. ChatGPT generates text based on data on the Internet, so it may involve the patient's personal information and privacy, and needs to be handled with more care.
• Interpretability of the model: ChatGPT is a black-box model that makes it difficult to explain why and how it generates text. In the medical field, the decisions and results of the model need to be interpreted, which can limit the application of ChatGPT.
• Not suitable for emergencies: In emergency situations, such as cardiac arrest, ChatGPT cannot provide timely diagnosis and treatment recommendations, which may affect the patient's life safety.

In addition, the advantage of ChatGPT is trainable, the disadvantage is also trainable, if people with ulterior motives use data bombs to train unregulated ChatGPT, it may induce it to give specific answers for commercialization. This is one of the important reasons why we should maintain a cautious attitude when we use ChatGPT in the future. The entry of ChatGPT into the medical field will also bring a series of important problems. For example, will the doctor's work be automated? Can AI replace doctors to provide medical advice? Will doctors eventually be replaced? However, the answers to these questions are not black and white. We need to look at the application of ChatGPT in the medical field through a scientific and objective perspective.

In the future, it is hoped that with the continuous iterative update of ChatGPT, it can inject more energy and bring more possibilities into the development of the medical field, so that artificial intelligence can truly help doctors and patients!

Bibliography:

Darkhabani M, Alrifaai MA, Elsalti A, Dvir YM, Mahroum N. ChatGPT and autoimmunity - A new weapon in the battlefield of knowledge. Autoimmun Rev. 2023 May 19;22(8):103360. doi: 10.1016/j.autrev.2023.103360. Epub ahead of print. PMID: 37211242.

ZHANG Daqing. Historical changes of artificial intelligence applied to medical services[J]. Medicine and Philosophy,2023,44(1):6-7.

ZHENG Minhua, XIE Zhongshi, HUANG Hua, et al. What will ChatGPT bring to the medical community[J]. Chinese Journal of Gastrointestinal Surgery,2023,26(3):248-252. DOI:10.3760/cma.j.cn441530-20230217-00040.