Artificial Competence

Abstract This article develops a virtue-epistemological distinction between general and narrow artificial intelligence (AI) systems by analysing AI competence through Ernest Sosa’s virtue-theoretic framework. First, I outline a general distinction between narrow and general AI. I then characterise AI in a way that bridges its terminology with virtue epistemological concepts. Drawing on Sosa’s idea that the telic structure is replicable for systems with merely functional-teleological aims, I argue that current AI systems can possess first-order competence. I argue that increasing domain-specific competence of narrow AI cannot elevate it to the status of general AI, as that requires the constitutional disposition to assess the aptness of one’s own first-order performances, a second-order competence that no amount of first-order improvement can grant. Finally, I argue that the virtue-theoretic distinction between artificial narrow intelligence and artificial general intelligence depends on whether the AI system possesses second-order constitutional competence.