INFORMATION TECHNOLOGY

Social intelligence – the next frontier of artificial intelligence


Siri and Google Assistant have been able to schedule meetings based on user requests, but they don’t currently have the social intelligence to prioritize things. Recent research suggests that the development of artificial intelligence (AI) may run into obstacles due to a lack of social intelligence.

On March 10, Zhu Songchun’s team published a review paper in CAAI Artificial Intelligence Research, calling for attention to the field of Artificial Social Intelligence (ASI).

“AI has changed our society and daily life,” said Dr. Fan Lifeng, first author of the paper and State Key Laboratory of General Artificial Intelligence at the Beijing Institute of General Artificial Intelligence (BIGAI). We see artificial social intelligence (ASI) as the next big frontier. ”

The researchers believe that ASI consists of several subfields, including social perception, theory of mind (the ability to understand one’s own mental state and that of those around them), and social interaction. Fan Lifeng pointed out that the use of cognitive science and computational modeling to determine the gap between artificial intelligence systems and human social intelligence, as well as current problems and future directions, will make the field better.

“Compared to our physical intelligence, ASI is unique and challenging, and it is highly dependent on the environment,” Fan said. “In this study, the meaning of context can be as large as the cultural background and general knowledge of the population, or as small as the shared experience of two friends. This span poses challenges that make it impossible for standard algorithms to solve real-world ASI problems, which are typically complex, fuzzy, dynamic, random, partially observable, and multi-agent. ”

Therefore, implementing ASI requires a more holistic approach, as improvements to specific local components often do not improve the overall performance of ASI as desired, unlike today’s AI systems. ASI needs to be able to parse latent information expressed by humans in social activities, such as rolling their eyes or yawning, in order to understand the other person’s psychological state, such as thoughts and intentions, in order to cooperate in completing the set task.

“Multidisciplinary research provides information and inspiration for ASI’s research, and studying human social intelligence helps us gain insight into the fundamentals, compendiums, reference systems, and test benchmarks needed to develop ASI with humanoid characteristics,” Fan said, “We focus on the three most important aspects of social intelligence, which are inextricably linked: social perception, theory of mind, and social interaction, as they are all based on well-established cognitive science theories and are ready-made tools for developing computational models in these fields.” ”

Fan believes that the best way is to comprehensively and holistically mimic the patterns of human interaction with others and the surrounding environment. This requires an open, interactive environment and thinking about how best to introduce humanoid bias into ASI models.

“In order to accelerate the future development of ASI, we propose to take a comprehensive and systematic approach to model learning in multiple ways like humans, such as lifelong learning, multi-task learning, single-sample/small-sample learning, meta-learning, etc.,” Fan said, “We need to define new problems, create new environments and datasets, establish new evaluation methods, and establish new computational models.” The ultimate goal is to have a high level of ASI in AI and improve human well-being with the help of artificial social intelligence. ”

The authors also include Professors Xu Manjie, Cao Zhihao and Zhu Songchun of the Beijing Institute of General Artificial Intelligence; and Professor Zhu Yixin of the Institute of Artificial Intelligence of Peking University.

This research was supported by the Science and Technology Innovation 2030 Project of the Ministry of Science and Technology and the Beijing Science and Technology New Star Program.

The above content is based on the interview conducted by the editorial department of this journal on the author team, and reviewed by the author team.

DOI

10.26599/AIR.2022.9150010

Original title

Artificial Social Intelligence: A Comparative and Holistic View

CAAI Research on Artificial Intelligence (CAAI Artificial Intelligence ReseachFounded in 2022, it is hosted by the Chinese Institute of Engineering Intelligence and Tsinghua University, and published by Tsinghua University Press. The journal publishes high-level original research and review articles on artificial intelligence and its application fields, providing a high-end academic exchange platform for researchers at home and abroad. The paper was published on the SciOpen platform on open access.

About SciOpen

SciOpen is an international digital publishing platform for scientific and technological journals independently developed by Tsinghua University Press and with independent intellectual property rights. The platform consists of seven subsystems: submission review system, production system, all-media release system, cluster portal system, mobile/social system, knowledge service system and operation management system. The platform integrates the rich experience of Tsinghua University Press in running international journals, and is committed to building an international open science infrastructure, providing customizable full-process services for universities, research institutes, academic associations, journals, etc. to establish new journals and develop journals.

Special statement: This article is reproduced only for the need to disseminate information, and does not mean to represent the views of this website or confirm the authenticity of its content; If other media, websites or individuals reprint and use from this website, they must retain the “source” indicated on this website and bear their own legal responsibilities such as copyright; If the author does not wish to be reprinted or contact the reprint fee, please contact us.



Source link

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button