Philosophical foresight on artificial intelligence

BY PAN YUEFEI | 06-15-2017
(Chinese Social Sciences Today)

At the China Go Wuzhen Summit held on May 23, 25 and 27 in Jiaxing, Zhejiang Province, the best human Go player Ke Jie was defeated by the artificial intelligence (AI) arm of Google AlphaGo with a score of 3-0.


AlphaGo’s victory over Ke Jie once again brought AI to the spotlight. As a new discipline with distinct characteristics, the study of AI needs to interact with other disciplines for further development. Philosophy provides an important angle through which research on AI can be promoted.

 

Philosophical reflection
Li Jianhui, a professor of philosophy from Beijing Normal University, said philosophical reflection is essential because philosophy helps AI researchers clarify problems, outline fields and extract perspectives, as well as analyze and comment on AI development.


Clarifying problems in this instance refers to helping AI experts propose adequate and meaningful questions while outlining fields refers to understanding and describing how AI has been properly used to form large-scale maps of concepts. In addition, philosophy provides AI research with a new perspective and methodology. Philosophy provides theoretical comment on the ideology and methods of AI development, and analyzes its advantages and disadvantages, thus promoting theoretical and practical progress on AI.


Duan Weiwen, a research fellow from the Institute of Philosophy at the Chinese Academy of Social Sciences, said the new AI fervor came amid a technological breakthrough. In the process, a series of key issues call for a conversation between science and philosophy. First, the most dominant epistemological paradox is that people still do not have a sophisticated understanding of the mechanisms of machine learning, namely, the algorithm black box. Second, in the various intelligent estimations and decision-making systems based on machine learning and data analysis, apart from cognitive opacity there is value opacity. Last but not least, AI development is comprehensively transforming human lifestyles and even redefining what it means to be human. Issues relating to these aspects are closely tied to philosophical questions like epistemology, axiology and human essence, which require the active participation and foresight of modern philosophy.

 

Interpretability
Emerging disciplines, represented by AI, are challenging traditional philosophical theories. New ideas formed in the process promote theoretical reform and innovation.


Duan said issues relating to three aspects are worthy of attention. The first is the epistemological transparency of intelligence algorithms. The key to the issue lies in the interpretability of the intelligent computing process. This involves developing a second-order cognitive tool that can explain intelligent algorithms and their computational processes. From a philosophical point of view, the discussion over the possibility of introducing such a tool involves an epistemological definition of what kind of explanation of intellectual cognition is acceptable. The second is social prejudice and discrimination triggered by intelligent algorithms, with algorithm power at the core. The third is data power and privacy. The difficulty here is in redefining individual privacy rights from a philosophical perspective in the age of digitization and intelligence.

 

Challenge
Currently, AI development in China has seen a rise in the number of theories, technologies and practical innovations. Duan said the philosophical reflection on AI needs the participation of science and technology, industry and politics, as well as reason and insights into AI technologies and industrial practices. He argued that, the most noteworthy frontier issue at the theoretical level is the consideration of AI cognitive processes and the ethics of intelligent decision-making. This involves taking into account AI interpretability and transparency in terms of cognition and value. At the practical level, the most important thing is to emphasize responsible innovation in the field of AI and robots. On the one hand, this involves moral and ethical challenges put forward by emerging technologies, while the other involves issues concerning technological development and social governance, such as future innovation orientation, industrial patterns and employment policy.


How can philosophy contribute to future AI research? Li thinks that in the future, the philosophical discussion of AI will focus on the theoretical problems of computationalism. As long as these problems are solved and applied in practice, AI will enter a new stage of development.


Xu Yingjin, a professor of philosophy from Fudan University, said that issues associated with natural language processing from the perspectives of existing cognitive linguistics and cognitive psychology are worth studying. At present, there is a trend in natural language processing, a blind faith that machine translation can be accomplished by in-depth studying and big data methods, while less attention has been paid to how humans deal with the real psychological semantics of natural language processing. In this process, philosophy plays a role of finding those machine accessible dimensions in the theories of psychological semantic processing. In addition, the blind worship of big data technology indicates that philosophical reflection on intelligence is still sorely needed.

 

 

PAN YUEFEI is a reporter at the Chinese Social Sciences Today.