A study of Chinese scientists is expected to make artificial intelligence more "understand" humans

A study led by Chinese scientists has made breakthroughs in allowing artificial intelligence to quickly identify human value intentions, and can improve the efficiency of human -machine collaboration in complex tasks. The research results have been published online by the international academic journal "Science · Robotics".

The study was completed by the team led by Zhu Songchun, a professor at the Beijing General Intelligence Research Institute. The research team put forward a cooperation framework for the two -way value of man -machine and machine, and shows how robots have completed a series of complex human -machine collaboration tasks with human users through real -time communication.

Scientists have always tried to allow artificial intelligence to learn human cognitive reasoning, and then learn to understand human intentions, and implement the value demand that humans really care about. However, the currently widely used artificial intelligence systems often can only act mechanically according to the task given by humans.

"If the goal of a task can be clearly described, it is not difficult to let the value demand behind artificial intelligence understanding the task." Zheng Zilong, co -author of the paper and researcher at the Beijing General Artificial Intelligence Research Institute Task of target. For example, a cup of coffee means different drinks in the hearts of different people. When human users say "I want a cup of coffee", artificial intelligence needs to figure out what coffee is, that is, "alignment value" with humans.

According to reports, the research team proposed an explanatory artificial intelligence system based on the real -time two -way value alignment model. In this system, a group of robots quickly infer the value goals of human users through real -time communication with human beings, and at the same time convey their decision -making process to users through "explanation", allowing users to understand the value basis for judging robots. In addition, the system can generate a easier understanding explanation by speculating the inherent value preferences of human users.

The Beijing General Artificial Intelligence Research Institute is a non -profit new R & D institution supported by the Beijing Municipal Government and the Ministry of Science and Technology. It is supported by the cooperation between Peking University, Tsinghua University and other units.

Reporter: Dong Ruifeng

- END -