What is intelligence, really?

We live in an era where “artificial intelligence” is everywhere—AI helps us write emails, translate languages, recommend songs, and even drive cars. It’s easy to assume that, because computers can generate text or recognize faces, we must have cracked the code of how intelligence works. But ask anyone who’s worked on AI long enough and you’ll hear the same thing: even our most advanced models are still leagues away from human-like understanding.

So, why is real intelligence so hard to define—and even harder to replicate?

To answer that, we need to turn not to the world of computer science, but to the brain itself. And there’s no better guide for this journey than Jeff Hawkins, the Silicon Valley engineer-turned-neuroscientist whose “A Thousand Brains” theory is quietly shaking up how we think about both minds and machines.

Meet Jeff Hawkins: tech pioneer turned brain theorist

If you had a Palm Pilot in the early 2000s, you’ve already used something Jeff Hawkins helped create. But while many of his peers doubled down on the tech industry, Hawkins took a different route—he became obsessed with a much bigger question:

What is intelligence, and how does the brain actually work?

After years of reading, building, and thinking, Hawkins co-founded Numenta, a research company devoted to understanding the neocortex—the folded, walnut-like surface of the brain that’s responsible for perception, language, and, most importantly, intelligence. His work is chronicled in his book, A Thousand Brains: A New Theory of Intelligence, which takes us on a journey into the weird, beautiful architecture of the mind.

The thousand brains theory—how our minds actually build knowledge

We often picture the brain as a “central processor,” a single control room making all the big decisions. Hawkins blows up this idea completely. According to him, the neocortex is built from thousands of repeating units, called “cortical columns.” Each of these columns is like a tiny brain, capable of learning models of the world on its own.

Let’s break it down:

1. Cortical columns: mini-brains within your brain

• Your neocortex contains around 150,000 cortical columns.
• Each column processes a small part of sensory input (vision, touch, sound, etc.), building its own version of reality.
• When you look at a coffee cup, columns in your brain aren’t just seeing it—they’re building and updating models of “coffee cup-ness” from every possible angle, every sensation.

2. Reference frames: mapping the world

• Each column creates what Hawkins calls a “reference frame”—a way to map objects and their features in space.
• For example, your brain can recognize a cup whether it’s upside down, half-hidden, or cold to the touch, because columns are constantly updating their reference frames based on new data.
• This is why you can recognize your friend’s voice in a noisy room, or identify your keys by touch in the dark.

3. Consensus: the secret sauce of intelligence

• Here’s where things get wild: no single column “knows” what’s out there. It’s only when thousands of columns share information and reach consensus that you get a clear, stable understanding.
• This distributed approach is what makes human intelligence robust and adaptable—we don’t rely on one input or model, but a democracy of mini-brains, each voting on reality.

Table 1: how the brain builds intelligence vs. how most AI works

Why Hawkins’ theory changes the AI conversation

The implications for neuroscience are huge, but Hawkins’ ideas go even further. They challenge the basic assumptions behind modern artificial intelligence.

Today’s AI: brilliant, but blind

Most leading AI systems—whether built by OpenAI, DeepMind, Anthropic, or others—work by training a single, gigantic model on oceans of data. These models are amazing at pattern recognition, but they lack the parallel, context-rich modeling that our brains use every day.

• Sequential Processing: AI typically processes one thing at a time, relying on past data to make predictions.
• Context Limits: Feed it a paragraph about Paris and then ask a question about Tokyo? Unless the context is explicit in the prompt, the AI may not “get” the shift.
• No Reference Frames: AI can recognize a dog in a photo, but might fail to “understand” the concept of “dog-ness” across senses, situations, or times of day.

In other words: today’s AI is powerful, but it’s not really intelligent in the human sense. It’s missing the messy, consensus-driven, context-rich intelligence that’s the hallmark of the neocortex.

The coffee cup test: why the brain still wins

Let’s come back to the coffee cup. If you hand it to a child—handle up, handle down, with or without coffee—the child will instantly know what it is, what it’s for, and how to use it. Their brain has built thousands of overlapping models that, together, make “coffee cup” an obvious reality.

Try the same with an AI model, and you’re more likely to get a best guess based on pixel patterns, text correlations, or statistical likelihoods. No model consensus, no reference frames, no built-in adaptability. The machine can’t know a cup the way you do.

Table 2: comparing intelligence—human vs. today’s AI

Why this matters—not just for AI, but for how we build the future

If you care about the future of artificial intelligence, Hawkins’ “thousand brains” theory is more than a neuroscience curiosity. It’s a blueprint for what comes next.

Imagine an AI that doesn’t just parrot information, but actually models reality the way your brain does:

• Multiple, independent models working in parallel
• Robust context, always updating as new information comes in
• Ability to build understanding from touch, sight, sound—all at once
• Resilience to missing or ambiguous data

This kind of system wouldn’t just answer questions—it would understand them. It could adapt, learn new skills quickly, and make sense of the world in a deeply human way.

So, why haven’t we built this yet?

There are a few reasons:

• Technical Complexity: Building thousands of interconnected models that share information in real time is hard—far more complex than training a single large model.
• Data Integration: Unlike brains, computers don’t natively combine sensory data from different sources.
• Consensus Algorithms: It’s challenging to design digital “columns” that can truly reach consensus the way the neocortex does.

But research is moving fast. More and more teams are realizing that context and distributed intelligence are keys to closing the gap.

Where does calk AI fit in?

At Calk AI, we’re fascinated by brain-inspired ideas that go beyond the limits of traditional artificial intelligence companies. While we don’t claim to have recreated the neocortex, we do believe that the best artificial intelligence—whether you’re building with custom GPTs, exploring open source artificial intelligence, or integrating the best AI apps—must be context-aware, flexible, and collaborative. That’s why our approach to how to use artificial intelligence in your business isn’t about siloed, rigid workflows or fixed “custom GPT” interfaces. Instead, we help users create AI agents that combine knowledge, share context, and continuously update their understanding as your business evolves.

Like the brain’s columns, our vision for how to use AI in business is to move away from fragmented solutions and toward systems that truly work together—each agent bringing its own “reference frame” to the table. Whether you’re an enterprise seeking to implement a new AI strategy, an AI startup looking for scalable agent orchestration, or simply asking how to use AI in my business, our tools are designed to make it easy to create an AI system that grows and adapts with you. We’re committed to showing you how to use AI in your business, and what’s possible when artificial intelligence stops being just a tool and starts being a partner.

The road ahead: from brains to bytes

The story of intelligence—both human and artificial—is still being written. Hawkins’ theory doesn’t give us all the answers, but it’s a much-needed reminder: true intelligence is less about raw power, and more about how we represent, combine, and update knowledge in context.

It’s about building a thousand models of the world, and finding meaning in their overlap.

That’s how our brains work—and maybe, just maybe, it’s how our AIs will work in the future.

As we move forward, let’s keep one eye on the incredible architecture inside our own heads. The brain has been running its intelligence “operating system” for millions of years; we’ve only just begun to learn from it.

‍