The machine has "heart" AI can "understand" human values

Robot and human users complete a series of complex human -machine collaboration tasks through real -time communication

Source ｜ Science and Technology Daily

Author ｜ Hua Ling Dong Yifei

Edit ｜ Zhang Runqi

The reporter learned from the Peking University Artificial Intelligence Research Institute on the 13th that the joint team led by the professor of the institute Zhu Songchun achieved a major research result to build a calculation framework for real -time understanding of human values ​​to allow robots and human users to communicate in real time through real -time communication. Complete a series of complex human -machine collaboration tasks. This study can be described as a "heart" of the machine, allowing AI (artificial intelligence) to empower the robot to "understand" human values. The relevant papers are published in the latest issue of "Real -time Two -way Human -Machine Value" on the latest issue of the international journal "Science · Robotics".

It is understood that the widely used AI system today is a passive intelligence. It can only act mechanically according to the task given by humans, and lacks the same cognition and reasoning ability, and emotion and values ​​like humans. In the case of lack of "heart", it is difficult for AI to understand the true intentions and value needs of human beings, and naturally it is difficult to obtain human trust and integrate into the society.

After discovering the limitations of the previous research paradigm, Zhu Songchun led the research team to convert the track, allowing the AI ​​system to independently define the task according to the man's way. , Let AI learn the value function of human beings, have the ability to align the current human value goals in real time, and explore the paradigm of "small data, big tasks".

The research team proposes a explanatory AI system based on the real -time two -way value alignment model, allowing the machine to infer the value goals of human users through the instant interaction with humans and pass the corresponding feedback, and convey its decision -making process to users through the "explanation". Based on the value basis for judging robots. The research results show that the learning model can improve human -machine collaboration efficiency in complex collaborative tasks, thereby improving human and mechanical letter relationships, realizing true autonomous intelligence, and going further on the road to general AI.

contact number:

(010) 82081166-6059

Breaking the mailbox:

[email protected]

China Net Technology

Pay attention to the various hotspots of the science and technology circle, covering character dynamics, corporate stories, black technology products, scientific knowledge ... This is the gathering place for technology enthusiasts.

- END -