A research team in the United States developed an artificial intelligence system called Coscientist, which recently autonomously learned Nobel Prize level chemical reactions for the first time and successfully designed experimental steps to complete the reaction within a few minutes. This means that artificial intelligence is expected to help scientists achieve scientific results faster and more in the future. The team led by researchers from Carnegie Mellon University stated in an article published in the British journal Nature that the core of Coscientist is the multimodal large-scale language model GPT-4, and its palladium catalyzed cross coupling reaction has been widely used in global scientific research, medicine, and electronics industries. In 2010, three scientists from the United States and Japan were awarded the Nobel Prize in Chemistry for their outstanding contributions in the field of this reaction. Researchers claim that Coscientist can search for public information about compounds on the internet, find and read technical manuals on how to control robot laboratory equipment, write computer code to conduct experiments, and analyze the validity of result data. In experimental operations, Coscientist can manipulate high-tech robots and other experimental equipment in the laboratory to accurately absorb, spray, heat, shake small liquid samples, and ultimately successfully synthesize the target chemical substance. Before completing the palladium catalyzed cross coupling reaction, the research team conducted separate tests on the different software modules equipped with Coscientist. In one of the tests, if the steps designed by Coscientist were followed, commonly used substances such as aspirin, acetaminophen, and ibuprofen could be successfully produced. Researchers say that for the first time, artificial intelligence has planned, designed, and executed complex chemical reactions invented by humans. In scientific research, there is a repetitive process of trial, failure, learning, and improvement, and artificial intelligence is expected to greatly accelerate this process because the latter can "think" 24/7, making up for the shortcomings of human scientists. (Lai Xin She)