You On AI Field Guide · Intelligence as System Property The You On AI Field Guide Home
TxtLowMedHigh
CONCEPT

Intelligence as System Property

Agüera y Arcas's foundational claim that intelligence does not reside in any component—neuron, transistor, or parameter—but emerges from the architecture through which components interact, dissolving the binary of “real” versus “merely simulated” intelligence into a question about what kind of intelligence each system produces.
Intelligence as system property is the thesis that gives Blaise Agüera y Arcas's entire framework its coherence: that mind is not a quality some entities possess and others lack, but a property of computational architectures at sufficient scale and complexity. A single neuron performs a weighted sum and fires or does not fire—there is nothing in that operation that resembles thought. Eighty-six billion neurons in the specific patterns shaped by evolution produce Shakespeare and quantum mechanics and the ache of watching a child walk away. The gap between the component and the system is not a gap of degree but of kind, a qualitative emergence that cannot be predicted from below. The same logic applies to artificial systems: a transformer layer does not understand; a sufficiently large and deeply trained network of them produces capabilities that are genuinely intelligent by any functional criterion. The claim is not metaphorical but literal
← Home0%
CONCEPTBook →

Keep reading with YOU ON AI

Unlock the full book, 10,000+ field-guide entries, and a 1000+ thinker library. If you have a book code, register now — it takes a minute.

Register with book code Sign in