Two critical points of the EU AI Act
Abstract. This paper puts forward two broad criticisms of the proposed EU AI Act. First, the proposal does not address the future developments of general intelligent systems and it is ill-equipped to deal with more general existing systems like GPT-3. Second, one suggestion found in the literature to meet the first criticism, i.e. to increase transparency, fails, and not only because the proposal itself fails to specify the transparency standards required, but because it is problematic to spell transparency out in a concrete, usable way.