大会シンポジウム「人工知能・ロボットの哲学」
人工知能について、哲学的に考えるために
言語理解という観点から
2019 年 2019 巻 70 号 p. 35-50
詳細
2019 年 2019 巻 70 号 p. 35-50
This paper describes some basic concepts recent trends in AI and natural language processing (NLP), as background knowledge to discuss the relationship between AI and philosophy. The driving force of the current AI boom is machine learning (ML), which has also been the mainstream technology in NLP since 1990s.
Although the problem of data shortage is a major barrier for ML, a lot of efforts have been made to obtain high accuracy while reducing the labor of data collection.
However, recent achievements in ML and NLP do not directly lead to the AI which “understands” human languages. ML is neither foolproof nor versatile, and even with the future development of learning algorithms, it may not be easy to reduce human efforts to guarantee the “well-definedness” of tasks, which is a prerequisite for ML. There are also extremely difficult tasks such as understanding of ‘speaker’s intention’, which needs complex inference with various types of knowledge.
There is no doubt that AI technology will continue to develop, and will bring about major changes in human society. In order to lead AI research to the right direction, philosophy should play a major role as a “guide.” Especially, I expect philosophers to reconsider/refine basic concepts such as “understanding” and “intelligence” based on the collaboration with AI researchers.
