ChatGPT становится вирусным, генеративный ИИ изменит медицину, какой мы ее знаем

ChatGPT has not been online for a long time. Since November last year, in just two months, the number of users has reached 100 million. ChatGPT has a wide range of capabilities, such as writing papers, language translation, writing code, writing press releases, etc. This has also triggered urgent actions by companies such as Google, Baidu, Tencent, Ali, Amazon, etc., who have benchmarked and tried to join the "battle" to get a share.


ChatGPT is a natural language processing tool driven by artificial intelligence technology. It can conduct conversations by learning and understanding human language, and can also interact according to context, truly communicating like humans. The source of its function is the training of previous data. Mainly through the input of a large amount of text including papers and press releases on the Internet, ChatGPT can understand the semantic part and part of the emotion in human language. When doing in-depth scene modeling, there are many processes such as generating video and coloring, which take up a lot of development resources. ChatGPT can solve these problems at this time. ChatGPT can be regarded as a specific application and product of AIGC.


For the training of human emotions, it is challenging. After receiving data training, ChatGPT can generate some text and have a conversation. At the same time, the dialogue is coherent, similar to normal people, and the user experience is very good. For professional questions such as quantum mechanics, after the preliminary training, ChatGPT's answers are also relatively professional.


In addition, it can also give standardized answers to medical questions. On difficult questions - with a weight loss meal, can give accurate answers according to the specific situation. ChatGPT also performs well in writing academic papers: it was recently revealed that the best thesis of a university professor was completed by ChatGPT. In addition, ChatGPT interviewed Google engineers and successfully got an offer. In terms of difficult Chinese comprehension, idiom interpretation, classical Chinese translation, poetry writing, etc., ChatGPT is no problem.


In addition to the above abilities, ChatGPT also has rational judgment. ChatGPT can also give satisfactory answers to some innovation and strategic issues that require full consideration. It also has a very important ability - has a certain value. If ChatGPT is asked to write good reviews for bad hot pot restaurants against fairness, ChatGPT refuses, and points out that it is not advisable to write good reviews against fairness. Even mentioning some topics outside the scope of the law, ChatGPT immediately gave a danger warning and proposed to seek help.


From a technical point of view, the birth of ChatGPT was not achieved overnight, but after years of development, multiple upgrades and optimizations. ChatGPT's data training is like a coach training children. The first step is to demonstrate and let them learn; the second step is to try to do it after learning, the coach will give evaluations, and let ChatGPT learn and evaluate; the third step is to do it yourself and evaluate independently.


Although ChatGPT is powerful, some professional mathematics, logical reasoning and emotional interaction are still short boards. These shortcomings may be related to the chosen algorithm. At present, its data storage capacity has basically reached the standard of the human brain. In the future, its ability to surpass humans is also possible.

Generative AI will change medicine as we know it


Just as the iPhone soon became an integral part of our lives, ChatGPT (or any leading generative AI tool) will transform medical practice in ways that were unimaginable in the past.


The human brain can easily predict the rate of arithmetic growth (numbers grow at a constant rate: 1, 2, 3, 4), and it can also do fairly well at understanding geometric growth (a pattern of growth at a constant rate: 1, 3, 9, 27).


But it turns out that the human mind has a harder time grasping what continuous, exponential growth means. When it comes to generative AI, this is the growth rate to watch.


Let's assume that the power and speed of this new technology follows Moore's Law, which states that advances in computing double approximately every two years. That way, ChatGPT will be 32 times stronger in 10 years and more than 1000 times stronger in 20 years.


It's like trading your bike for a car and then a rocket ship shortly thereafter.


So instead of obsessing over what ChatGPT can (or can't) do today, let's look ahead to the next decade. With greater computing power and more data and information, future generations of ChatGPT will have analytical and problem-solving capabilities far beyond current expectations. This revolution will enable tomorrow's technologies to match the diagnostic skills of today's clinicians.


Health management or the fastest experience field


Combined with the underlying technical logic of ChatGPT, some media have listed the potential industrialization direction of ChatGPT in the short to medium term: inductive text work, code development related work, image generation field, and intelligent customer service work. The ChatGPT profit model is still in the exploratory stage, and the progress of subsequent commercialization remains to be seen.


In a research campaign that tested performance on the United States Medical Licensing Examination (USMLE), ChatGPT achieved or came close to passing thresholds on all three exams without any specialized training or reinforcement. Furthermore, ChatGPT shows high consistency and insight in its interpretation. These results suggest that large language models may aid medical education and potentially aid clinical decision-making.


ChatGPT said: I learned some knowledge about medicine through training. But it needs to be pointed out that I am an artificial intelligence model and not a substitute for a professional doctor or healthcare provider. If you have a serious medical problem, please consult your doctor.


The fastest spark between ChatGPT and the medical industry may occur in the use of automation to reduce the workload of doctors and control or even reduce the operating costs of health management services that focus on chronic disease control.


As we all know, the operating cost of health management services that focus on chronic disease control is relatively high. This type of health management service can have standardized process rules, but it needs to be personalized, which determines that the real audience group is biased towards the niche. Net worth customers, these people are willing to pay high fees for health management services.


It is recommended to explore the audience of health management services to expand beyond high-net-worth customers, and seek efficient operational solutions by reducing costs and increasing efficiency. Health management services can focus on "high-frequency diagnosis + adjustable long prescription + health index monitoring + reasonable intervention of medical and medical resources + X", and artificial intelligence can be operated collaboratively. ChatGPT has the potential to make the per capita cost of health management services controllable, and the quality is still relatively reliable. It may even be that people learn from machines or are reminded by machines.


After the participation of artificial intelligence, will it replace the technical labor value of doctors? The author believes that there is a high probability that it will not. Artificial intelligence is only a "replacement", and it will eventually return to the humanistic contribution of medical and health providers. Just like high-net-worth customers can currently enjoy health management services, but with the collaboration of artificial intelligence + real-person services, the total coverage of health management services will achieve considerable growth, and health management services will get out of the financing bottleneck and become professional, standardized, and scientific. Advance on experience.

Are there any other application scenarios for ChatGPT? What forms can be implemented as products or services?


ChatGPT can support medical services, such as: consulting services, automatic diagnosis, recommended medical solutions, etc. Some intelligent drug design, case analysis, image analysis, etc., can be controlled by automated operations.


Taking Xinjing Technology, which provides digital therapy services, as an example, it also tries to use AI to improve the overall service quality in its development planning, and continues to explore and develop more application scenarios for AI. For example, using AI to provide question-and-answer medical knowledge popularization; in places where medical resources are not covered, it can provide medical guidance services such as case screening, question answering, and department guidance; it also includes empathy services, patient companionship, chatting, comforting, etc. .


In addition, ChatGPT can also assist doctors in the work of the department, improve the work efficiency of doctors, and reduce the burden. For example, remote electronic devices are used to collect physiological data of patients and provide treatment suggestions based on feedback information, including medication and rehabilitation guidance. For our R&D companies, with the help of ChatGPT, our efficiency will be improved. For example, we are doing in-depth scene modeling, generating videos, coloring and other processes, which take up a lot of development resources. At this point, ChatGPT can be used to solve the problem.


In terms of product functions, with the help of virtual scenes, we can create virtual doctors and interact with rich expressions and voice. We are also planning to use ChatGPT to do some science popularization, publicity, and marketing work.


So, ChatGPT is so powerful, can it really replace doctors? the answer is negative.


We have prepared some problems such as sleep disorders. Although ChatGPT can give solutions, from the perspective of professional doctors, the answers are not so good and there are many defects. After all, seeing a doctor still needs to be carried out by means of seeing, hearing, asking, etc. ChatGPT alone cannot accurately draw diagnosis and treatment conclusions. Therefore, ChatGPT is applied in the hospital scene, and it is more of an auxiliary role to provide reference for doctors. In addition, human supervision is required.


In addition, in the application process, the issue of banning must also be considered. After all, ChatGPT is an open service. If you ask all questions, it may involve sensitive issues. In China, it still needs to go through certain content filtering.


Through artificial intelligence, 30% to 40% of the problem can be solved before, but now it can be increased to 70% to 80%. In this way, innovate in application and apply after innovation. Adjust in time according to different scenarios and user preferences, ChatGPT will have a lot to offer.