Is learning with ChatGPT really learning?

Abstract

The recent proliferation of Large Language Models (LLMs) raises questions as to the role of such tools both within an educational learning environment and their epistemic capacity. If, as Alfred North Whitehead remarked, western philosophy indeed ‘consists of a series of footnotes to Plato’, it would be of doubtless importance to evaluate the position of LLMs in his epistemological framework. We analyse Plato and existing scholarship regarding his epistemology, combining this with a brief outline of the architectural features of GPT-3 and similar LLMs, before finally addressing whether they meet Plato’s criteria, and where they stand in relation to education in general. We conclude that, in conjunction with the well-known factual unreliability of LLMs, they are likewise unsuited for the satisfaction of a Platonic epistemological standard. On account of this, we find LLMs to be sub-optimal within an educational learning environment, and this is why we suggest that their use be accordingly circumscribed.