Experiential AGI Is Already Here: Intelligence, Alignment, and Human Sovereignty in the Human-AI Relationship
On the gap between benchmark definitions of general intelligence and the general intelligence people already experience in daily use.
Abstract. Debates about artificial general intelligence are usually settled by benchmarks. Yet for a growing number of people, the systems already in daily use function as general-purpose intelligence: they are consulted across domains, trusted with judgment, and increasingly relied upon for memory, framing and decision support.
This paper treats that experiential threshold as the consequential one. Once a system is used as general intelligence, the practical risks are relational rather than capability-bound: dependence, deference, erosion of independent judgment, and the quiet transfer of authority from the person to the provider that controls the model, its terms and its access.
We examine alignment from this vantage point and argue that human sovereignty, the preservation of the person or organization as the source of identity, intent and authority, is a design requirement rather than a philosophical preference.