From a technical point of view, China’s ai is still in the weak ai stage.Although the benefit from the image recognition, deep learning, breakthroughs in key technologies such as neural network, the AI in the robot, speech recognition, remote control and planning, virtual personal assistant, medical and other fields has been widely used, but for most of the dividend policy of medical AI company, technical force is limiting the further development of the main obstacles.On the one hand, these companies still have technical bottlenecks for complex or multidisciplinary joint diagnosis algorithms, and their independent research and development and innovation capabilities need to be further improved.On the other hand, at present, China lacks a security assessment system and enterprises do not take sufficient measures to protect the privacy of medical data.
The quality and quantity of data is at the heart of the current competition for medical ai in China.However, for Chinese medical artificial intelligence enterprises, there are large-scale potential data in the market, but they cannot be sorted out and utilized.On the one hand, the number of hospitals in China is huge, but more than 75% of them are unstructured, which cannot bring the value of « big data » mining into play.On the other hand, both modeling and training machines are inseparable from the real clinical environment, which is currently lacking in most of the medical artificial intelligence products in China.
At the same time, the error of data will also cause obstacles to the development of artificial intelligence.In China’s current medical system, hospitals are not connected with hospitals and departments of internal medicine, there is no unified standard for clinical structured medical records, doctors’ handwritten medical records are not standardized, details such as clinical medication and examination are missing, and patients’ loss of access rate after leaving the hospital is high, which leads to the « accidental release » of health medical data.While deep learning needs to use large-scale standardized data for training, and subtle data errors will bring negative effects for deep learning.Such data quality inevitably raises doubts about the results of current medical artificial intelligence.
In addition to the challenges posed by technology and data, artificial intelligence faces ethical controversies.Can machine intelligence be the subject of ethics?In fact, although medical artificial intelligence brings a lot of convenience to medical diagnosis, treatment and rehabilitation, when faced with complex disease diagnosis and treatment, the « life and death power » cannot be handed over to artificial intelligence.If we completely rely on artificial intelligence, then the identification of liability for medical accidents and the identification of the regulatory liability for medical safety is a big problem.
previously, xu yingjin, a professor at the school of philosophy at fudan university, said that although robots are far from intelligent enough to become ethical subjects, it is not impossible for robots to have ethical awareness.To achieve this goal, it is necessary to study the relationship between ethical awareness and procedures, and to express the rules in a set of programming languages.