Artificial intelligence (AI) defends human health and well-being in a unique way. In addition to playing a role in the fields of diagnosis and treatment, medical management, and medical insurance, algorithm-based AI has promoted the innovation of drug research in recent years, and has increasingly demonstrated its advantages. Generally, the development of a drug can be divided into two stages: drug discovery and clinical research. In the drug discovery stage, scientists need to establish disease hypotheses, discover targets, design compounds, and then start preclinical research. However, traditional pharmaceutical companies must conduct a large number of simulation tests during the drug development process, which has a long development cycle, high cost and low success rate. According to Nature data, the research and development cost of a new drug is about 2.6 billion U.S. dollars, and it takes about 10 years. However, the output is less than one-tenth. Obstacles exist in every stage. For example, in discovering targets and designing compounds, including the screening of seed compounds, optimization of lead compounds, identification of candidate compounds, synthesis, etc, there could be many challenges. For the discovery of targets (, continuous experimental screening is required to find chemical molecules with therapeutic effects from hundreds of molecules. Even thousands of compounds may need to be screened for the development of one drug. Even so, there are only a few that can successfully enter the final R&D stage. However, through AI technology, it is possible to find in-depth connections between diseases, genes and drugs to reduce the high R&D costs and failure rate. Based on disease metabolism data, large-scale genome recognition, proteomics, and metabolomics, AI can perform virtual high-throughput screening of candidate compounds, find the relationship between drugs and diseases, diseases and genes, improve drug development efficiency as well as the success rate of drug development. Specifically, researchers can use the text analysis function of AI to search and analyze massive literature, patents, and clinical results, find potential and neglected pathways, proteins, mechanisms, and other related relationships with diseases, and further propose new possibilities or hypothesis for testing, so as to find new mechanisms and new targets. Amyotrophic lateral sclerosis (ALS) is a rare disease caused by specific genes, and IBM Watson uses AI technology to detect the association of tens of thousands of genes with ALS, and successfully discovered 5 genes related to ALS, which has advanced progress in human research on ALS (previously, 3 genes related to ALS have been discovered in medicine). In terms of candidate compounds, AI can perform virtual screening (, helping researchers to efficiently find compounds with higher activity, and improve the screening speed and success rate of potential drugs. For example, with MedAI’s structure-based drug design platform, drug database can be swiftly analyzed to simulate the development process, predict potential drug candidates, evaluate new drug development risks, and predict drug effects. The company can even use neural network-based algorithms to find new drug candidates and predict disease biomarkers. When drug development has gone through the drug discovery stage and successfully entered the clinical research stage, it has entered the most time-consuming and costly stage of the entire drug approval process. Clinical trials are divided into multiple phases, including clinical phase I (safety), clinical phase II (effectiveness), and clinical phase III (large-scale safety and effectiveness) testing. In traditional clinical trials, the cost of recruiting patients is high. A survey by CB Insights showed that the biggest reason for the delay of clinical trials was the recruitment process. About 80% of trials failed to find the ideal test drug volunteers on time. An important part of clinical trials is to strictly abide by the agreement. In short, if the volunteer fails to comply with the experimental rules, then the relevant data must be deleted from the collection. Otherwise, if it is not discovered in time, these data containing the wrong medication background may seriously distort the test results. In addition, ensuring that participants take the right medicine at the right time is equally important for maintaining the accuracy of the results. But these difficulties can be solved by using AI technology. For example, AI can use technology to extract effective information from patient medical records and match it with ongoing clinical research, which greatly simplifies the recruitment process. For the problems that exist in the experiment, such as the inability to monitor the patient`s medication compliance, AI technology can achieve continuous monitoring of patients, such as using sensors to track medication intake, and image and facial recognition to track patient medication compliance.