How Does Artificial Intelligence Mimic the Brain?

The Biological Blueprint of Deep Learning (0)

Background: The brain has three main parts: the Cerebrum, Cerebellum, and Brainstem. The Neocortex is located at the very outer edge of the Cerebrum. In terms of human brain evolution, the brain developed the Neocortex to make faster evolutionary progress. This allowed the brain to devise a way to create new behaviors and solutions without waiting for the older structures of the brain to genetically reconfigure themselves. (1)

Because of this fortunate evolutionary occurrence, our thinking process evolved to be incredibly flexible, allowing humans to invent new solutions to complex problems quickly. "Rather than being a collection of different modules controlling different behaviors," the Neocortex works like a collective whole. It fundamentally changes how we learn, making the process both better and faster. (2)

To achieve this massive leap in capability, the Neocortex grew larger and more efficient by developing physical folds in its anatomy, allowing it to pack massive surface area and distinct functional layers into a fixed structural space --- the skull.

The Power of the Column

The architecture of this structure allows it to efficiently compute certain types of repetitive activity. This uniform, highly organized structure relies on interconnected cortical columns to process complex information. While an individual column allows a localized "thinking process" to occur with roughly one hundred thousand neurons, (3) this architecture scales seamlessly across billions of cells. It transforms raw sensory inputs into abstract thoughts, predictions, and actions. (4)

Artificial intelligence tries to mimic that exact biological process by relying on Artificial Neural Networks. Today, we see examples of this all over the world: a single foundational network architecture can be trained to write poetry, analyze complex medical images, or drive an autonomous vehicle.

The Neocortex's reliance on a highly efficient, repetitive anatomical structure is exactly how Deep Learning works through large language module layering. (Say that fast five times.)

Instead of writing custom code for every single unique problem, AI engineers stack identical mathematical layers—layers of artificial neurons—on top of one another. They let the system organize its own "thinking process" and build increasingly abstract representations as data moves through the stack.

Expanding Capacity: Folds vs. Parameters (5)

Just as the biological brain evolved physical folds to pack more computing power into a fixed skull space, AI scales up its capacity digitally. It increases its "parameter count"—the hundreds of billions of digital connections within its network—to handle increasingly complex data. (6)

When an AI model provides a different or updated answer months after a previous query, it isn't because the layers are rewiring their own code on the fly; rather, it reflects updates to the data, fine-tuning adjustments, or live web retrieval parameters integrated into those layers by engineers.

This is just a brief look at the fascinating bridge between organic evolution and digital architecture, but it highlights a profound truth: the most advanced future we are building is deeply rooted in the biological structures we've carried for millennia.

________________________________________________________

Notes & References

(0) At its core, Deep Learning is a subset of artificial intelligence (AI) and machine learning (ML) that mimics the way the human brain processes data and creates patterns for use in decision-making. The "deep" in deep learning specifically refers to the structure of the software: it uses Artificial Neural Networks (ANNs) with many layers stacked on top of one another. (https://gemini.google.com/app/739593af80a168aa)

(1) Ray Kurzweil, "The Singularity is Nearer", Viking Press, 2024, page 33.

(2) Ibid, page 34.
(3) Ibid.
(4) Ibid.
(5) The phrase "contextual parameters" in this framework refers to the situational boundaries, background rules, or environmental cues that the brain (or an AI) uses to make sense of information. In a truly cognitive sense, a piece of data never exists in a vacuum. The neocortex doesn't just process raw input; it filters that input through a set of parameters established by the surrounding context.
(6) Verified via architectural frameworks of modern transformer networks.