Zagadnienia Filozoficzne w Nauce (Dec 2022)

meta-ontology of AI systems with human-level intelligence

  • Roman Krzanowski,
  • Pawel Polak

Journal volume & issue
Vol. 73
pp. 197 – 230

Abstract

Read online

In this paper, we examine the meta-ontology of AI systems with human-level intelligence, with us denoting such AI systems as AIE. Meta-ontology in philosophy is a discourse centered on ontology, ontological commitment, and the truth condition of ontological theories. We therefore discuss how meta-ontology is conceptualized for AIE systems. We posit that the meta-ontology of AIE systems is not concerned with computational representations of reality in the form of structures, data constructs, or computational concepts, while the ontological commitment of AIE systems is directed toward what exists in the outside world. Furthermore, the truth condition of the ontology (which is meta-ontological assumption) of AIE systems does not require consistency with closed conceptual schema or ontological theories but rather with reality, or in other words, “what is the world” (Smith, 2019, p.57). In addition, the truth condition of AIE systems is verified through operational success rather than by coherence with theories. This work builds on ontological postulates about AI systems that were formulated by Brian Cantwell Smith (2019).

Keywords