DEEP LEARNING IN SCIENCE: IS THERE A REASON FOR (PHILOSOPHICAL) PESSIMISM?

In this article, I will review existing arguments for and against this philosophical pessimism about using deep learning models in science. Despite the remarkable results achieved by deep learning models networks in various scientific fields, some philosophers worry that because of their opacity, using these systems cannot improve our understanding of the phenomena studied. First, some terminological and conceptual clarification is provided. Then, I present a case for optimism, arguing that using opaque models does not hinder the possibility of gaining new understanding. After that, I present a critique of this argument. Finally, I present a case for pessimism, concluding that there are reasons to be pessimistic about the ability of deep learning models to provide us with new understanding of phenomena, studied by scientists.