Por Mariana Seabra (FLUL).

A considerable part of current AI is based on pattern recognition algorithms. From a philosophical point of view, it becomes imperative to ask i) is it possible to define pattern? and ii) do the patterns in question (i.e. handled by AI algorithms) actually correspond to structures in the world, or conversely, impose a certain description of the world? I argue that a satisfactory answer to the aforementioned questions requires a systematized analysis of the assumptions behind each particular pattern recognition method and also knowledge on crucial technical details. Therefore, in this talk I discuss these matters for a selection of pattern recognition algorithms. Furthermore, I try to establish their epistemic limits in terms of both explanatory power they may have and scientific knowledge and justification they may deliver.

Transmissão via Zoom (pw: 270389).