Paul D’Ambrosio’s article “What Philosophy Can Learn from AI: Methodological Insights from Deep Learning and Chinese Philosophy” has just been published by Synthese.

Abstract:

As artificial intelligence exerts growing influence across society, academic philosophers are increasingly called upon to address questions concerning ethics, alignment, agency, and meaning. Yet recent developments in AI can also invite reflection on certain prominent philosophical methodologies as well. Appreciating the success of deep learning systems raises questions about the degree to which effective understanding depends upon explicit definitions, conceptual clarification, and linear forms of reasoning. It also invites us to seriously reconsider the value of examples, analogies, patterns, tacit knowledge, and context-sensitive responsiveness in understanding and interacting in the world. This paper develops a limited structural comparison between selected methodological tendencies in contemporary Anglophone academic philosophy and supervised classical machine learning. For example, although the latter is statistical and driven by data, both tend to rely comparatively heavily upon predefined categories, predetermined structures, human-selected features, and relatively transparent forms of reasoning. By contrast, deep learning and selected approaches in early Confucian and Daoist philosophies exhibit noteworthy similarities in their reliance upon examples, cultivation, and iterative engagement with experience. While not equivalent, both do illustrate how productive patterns of response can be developed through repeated encounters with particulars, while relying less on pregiven specified conditions. We therefore argue that the success of deep learning invites renewed reflection on certain academic assumptions that effective forms of understanding and competence must depend heavily upon explicit conceptual articulation. In sum, the paper suggests that recent developments in AI provide an opportunity to reconsider the methodological scope of philosophical inquiry itself. We may also make space for philosophical approaches where inquiry begins with a variety of provisional models, examples, doctrines, or patterns, and then tests them, and iteratively revises them, in actual cases, according to specific experiences and contexts.

The full article is here.