自然杂志 ›› 2019, Vol. 41 ›› Issue (2): 79-84.doi: 10.3969/j.issn.0253-9608.2019.02.001

• 特约专稿 •    下一篇

踏踏实实地研究人工智能

郭毅可   

  1. 上海大学 计算机学院,上海 200444;英国帝国理工学院 数据科学研究所,英国 伦敦 SW7 2AZ
  • 收稿日期:2019-03-10 出版日期:2019-04-25 发布日期:2019-04-22
  • 作者简介:通信作者,欧洲科学院院士、英国皇家工程院院士。研究方向:人工智能、机器学习和数据科学。

Developing artificial intelligence with a down-to-earth approach

GUO Yike   

  1. School of Computer Science, Shanghai University, Shanghai 200444, China; Data Science Institute, Imperial College London, South Kensington Campus, London SW7 2AZ, UK
  • Received:2019-03-10 Online:2019-04-25 Published:2019-04-22

摘要:

今天人工智能的巨大成就表现在机器学习上取得了突破性的发展及“智能+”推动了人工智能的普适应用。人工智能成为世界科技发展的一个新高地,各国对此都作出战略布局。同时,人工智能的发展也向我们提出了新的挑战,在伦理、社会治理等方面引入了新的课题。在展望人工智能光明前景的同时,我们也要清楚地认识到:人工智能,特别是机器学习,它基本的方法、基本的思路还是比较简单和粗糙的。现在的人工智能是着重于智能外延的人工智能,也就是说着重于模拟人的智能的外在功能,而人工智能的发展还有待于在智能内涵的理解上的不断进展。本文就此对人工智能的发展作一个抛砖引玉的讨论, 也对机器学习的研究方向作探讨。

关键词: 人工智能, 机器学习, 数据科学

Abstract:

Today, artificial intelligence has been significant progress. The breakthrough development of machine learning has promoted the adaptation of artificial intelligence with a wide range of applications. Artificial intelligence become a new highland for the development of science and technology in the world, and all countries have made strategic investment. At the same time, the development of artificial intelligence has also presented us with a new challenge, introducing new topics in our ethics and social governance. While looking forward to the bright future of artificial intelligence, we must also clearly recognize the limitation of current state-of-the-art in artificial intelligence, especially machine learning. The basic methods and basic ideas are relatively still simple and rough. Today’s artificial intelligence is mainly focusing on the emulation of external function of the human intelligence. The development of artificial intelligence still needs to be continuously progressed in the understanding of the intelligent connotation. This article addresses the approaches for future development of artificial intelligence, especially the direction of machine learning research by emphasising a knowledge support and data driven methodology.