On Intelligence: Summary & Key Insights

The pursuit of artificial intelligence has a history rich in ambition but marked by misconception. Early AI pioneers believed that symbol manipulation—writing rules and logical statements—could encapsulate intelligence. They created systems that excelled in narrow domains, but these systems never generalized—they could not transfer learning from one situation to another or operate outside preprogrammed structures. They lacked something essential: context and adaptability.

At the same time, neuroscience advanced in isolating small pieces of brain function—neurons, synapses, neurotransmitters—but that reductionist approach also fell short of explaining intelligence as a whole. We could describe how individual neurons fire but not how billions of them together create perception, thought, or imagination.

Both fields were, in a sense, approaching the problem sideways—AI from computation without understanding, neuroscience from biology without synthesis. What was missing was an overarching *theory of intelligence* that could unite the two. That was the gap I set out to fill: to describe, at a systems level, how the brain’s architecture gives rise to intelligent behavior, and how that architecture could be emulated.

If there is a single structure in the brain that defines intelligence, it is the neocortex. This thin, folded sheet covering the top of the brain is remarkably uniform in structure across species that possess it. Whether it processes auditory information, visual scenes, or abstract thought, the same pattern repeats. That fact suggests something revolutionary: the neocortex does not care *what* information it handles—only *how* it handles it.

To understand intelligence, then, we must understand what the neocortex does. It receives a stream of sensory inputs, learns the regularities within those inputs, and uses them to make predictions about the future. If you think about walking through a familiar room in the dark, you know where the furniture is even without seeing it. That memory-guided anticipation is the neocortex at work, continuously predicting based on learned structure. Every sensation, perception, and thought arises from that predictive process.

This suggests that intelligence is a universal function—a set of operations performed by the neocortex—rather than a property of any particular brain region. It processes information hierarchically, builds models of the world, and constantly updates them as new evidence arrives. Intelligence, in short, is the ability to *predict the present from the past to shape the future*.