AI Prodigies — Envisioning a Different Future for Artificial Intelligence

This essay challenges the prevailing assumption that the goal of artificial intelligence should be increasingly general intelligence. It proposes an alternative model: “AI prodigies” — systems engineered for superhuman capability within deliberately bounded domains while remaining intentionally incomplete outside them. Using examples from AI safety research and the analogy of human prodigies, the essay asks whether superhuman performance can be achieved without creating broadly capable general agents. It also considers the strongest objection to this idea: whether sufficiently advanced reasoning in one field will inevitably transfer into others. The essay concludes by outlining the societal implications of specialized superintelligence, including the possibility that humans could retain general authority while relying on highly capable domain-specific AI systems.