Artificial Intelligence: The implications of current technological developments for Harry Collins epistemological theory and his critique of the possibilities of a general AI.

Abstract

Social constructivism and Harry Collins epistemological theory put focus on the sociology of scientific knowledge, scientific practice and the nature of expertise. From his theory, inspired by Ludwig Wittgenstein’s ‘forms of life’ and David Bloor’s strong programme, Collins has from the nineties until today argued against the possibilities of creating a general artificial intelligence, which can be compared with any cognitive functions that a human accommodates. Collins epistemological theory withholds aspects of knowledge basics, transfer and communication that are said to be unique and/or can only be performed by humans, which together with the problem of socialization constitute the demarcation line between a human and what can be achieved by an artificial intelligence, AI. Two areas of AI-research that in recent years have reported progress are autonomous vehicles and games. Autonomous vehicles are supposed to, in the near future, be able to handle different types of environments, interpret conflicting signals and ‘make’ ethical decisions. Game-playing AIs are said to be able to create their own set of rules and perform a basic communication. If these advancements are correct the AIs seem to challenge Collins critique of a potential general AI by exhibiting social sensibility, being able to adapt to different social and/or cultural settings and improvise when needed plus developing a (limited) language in order to communicate. In the paper I clarify what is at stake in the AI-debate, and intend to empirical test Harry Collins’ theories, and if his epistemological theory is affected by the reported and anticipated progress in the AI-research. My findings are that Collins’ epistemological theory is severely affected due to him, in his argumentation, slipping towards naturalism in order to construct a solid demarcation line, safeguarding against moving goalposts and a resistance to acknowledge the possibility of an AI being able to absorb anything equivalent to human social experience.