A Thousand Brains: A New Theory of Intelligence

by Jeff Hawkins

I find it amazing that the only thing in the universe that knows the universe exists is the three-pound mass of cells floating in our heads.

The evidence we have indicates that the complex circuitry seen everywhere in the neocortex performs a sensory-motor task. There are no pure motor regions and no pure sensory regions.

There can be little doubt of the dominating influence of the Darwinian revolution of the mid-nineteenth century upon concepts of the structure and function of the nervous system. The ideas of Spencer and Jackson and Sherrington and the many who followed them were rooted in the evolutionary theory that the brain develops in phylogeny by the successive addition of more cephalad parts. On this theory each new addition or enlargement was accompanied by the elaboration of more complex behavior, and at the same time, imposed a regulation upon more caudal and more primitive parts and the presumably more primitive behavior they control. What Mountcastle says in these first three sentences is that the brain grew large over evolutionary time by adding new brain parts on top of old brain parts. The older parts control more primitive behaviors while the newer parts create more sophisticated ones. Hopefully this sounds familiar, as I discussed this idea in the previous chapter. However, Mountcastle goes on to say that while much of the brain got bigger by adding new parts on top of old parts, that is not how the neocortex grew to occupy 70 percent of our brain. The neocortex got big by making many copies of the same thing: a basic circuit. Imagine watching a video of our brain evolving. The brain starts small. A new piece appears at one end, then another piece appears on top of that, and then another piece is appended on top of the previous pieces. At some point, millions of years ago,...

Recall that the neocortex is divided into dozens of regions, each of which performs a different function. If you look at the neocortex from the outside, you can’t see the regions; there are no demarcations, just like a satellite image doesn’t reveal political borders between countries. If you cut through the neocortex, you see a complex and detailed architecture. However, the details look similar no matter what region of the cortex you cut into. A slice of cortex responsible for vision looks like a slice of cortex responsible for touch, which looks like a slice of cortex responsible for language. Mountcastle proposed that the reason the regions look similar is that they are all doing the same thing. What makes them different is not their intrinsic function but what they are connected to. If you connect a cortical region to eyes, you get vision; if you connect the same cortical region to ears, you get hearing; and if you connect regions to other regions, you get higher thought, such as language. Mountcastle then points out that if we can discover the basic function of any part of the neocortex, we will understand how the entire thing works.

Looking at the surface of the neocortex, a cortical column occupies about one square millimeter. It extends through the entire 2.5 mm thickness, giving it a volume of 2.5 cubic millimeters. By this definition, there are roughly 150,000 cortical columns stacked side by side in a human neocortex. You can imagine a cortical column is like a little piece of thin spaghetti. A human neocortex is like 150,000 short pieces of spaghetti stacked vertically next to each other.

Mountcastle’s proposal looms in neuroscience like a holy grail. No matter what animal or what part of the brain a neuroscientist studies, somewhere, overtly or covertly, almost all neuroscientists want to understand how the human brain works. And that means understanding how the neocortex works. And that requires understanding what a cortical column does. In the end, our quest to understand the brain, our quest to understand intelligence, boils down to figuring out what a cortical column does and how it does it. Cortical columns are not the only mystery of the brain or the only mystery related to the neocortex. But understanding the cortical column is by far the largest and most important piece of the puzzle.

The brain creates a predictive model. This just means that the brain continuously predicts what its inputs will be. Prediction isn’t something that the brain does every now and then; it is an intrinsic property that never stops, and it serves an essential role in learning. When the brain’s predictions are verified, that means the brain’s model of the world is accurate. A mis-prediction causes you to attend to the error and update the model.

The big insight I had was that dendrite spikes are predictions. A dendrite spike occurs when a set of synapses close to each other on a distal dendrite get input at the same time, and it means that the neuron has recognized a pattern of activity in some other neurons. When the pattern of activity is detected, it creates a dendrite spike, which raises the voltage at the cell body, putting the cell into what we call a predictive state. The neuron is then primed to spike. It is similar to how a runner who hears “Ready, set…” is primed to start running. If a neuron in a predictive state subsequently gets enough proximal input to create an action potential spike, then the cell spikes a little bit sooner than it would have if the neuron was not in a predictive state.

In each minicolumn, multiple neurons respond to the same input pattern. They are like the runners on the starting line, all waiting for the same signal. If their preferred input arrives, they all want to start spiking. However, if one or more of the neurons are in the predictive state, our theory says, only those neurons spike and the other neurons are inhibited. Thus, when an input arrives that is unexpected, multiple neurons fire at once. If the input is predicted, then only the predictive-state neurons become active. This is a common observation about the neocortex: unexpected inputs cause a lot more activity than expected ones.

If all knowledge is stored this way, then what we commonly call thinking is actually moving through a space, through a reference frame. Your current thought, the thing that is in your head at any moment, is determined by the current location in the reference frame. As the location changes, the items stored at each location are recalled one at a time. Our thoughts are continually changing, but they are not random. What we think next depends on which direction we mentally move through a reference frame, in the same way that what we see next in a town depends on which direction we move from our current location.

What about destroying an intelligent machine when it is unplugged, or just never plugging it in again? Wouldn’t that be akin to murdering a person while they are asleep? Not really. Our fear of death is created by the older parts of our brain. If we detect a life-threatening situation, then the old brain creates the sensation of fear and we start acting in more reflexive ways. When we lose someone close to us, we mourn and feel sad. Fears and emotions are created by neurons in the old brain when they release hormones and other chemicals into the body. The neocortex may help the old brain decide when to release these chemicals, but without the old brain we would not sense fear or sadness. Fear of death and sorrow for loss are not required ingredients for a machine to be conscious or intelligent. Unless we go out of our way to give machines equivalent fears and emotions, they will not care at all if they are shut down, disassembled, or scrapped.

I expect that a similar change of attitude will occur with consciousness. At some point in the future, we will accept that any system that learns a model of the world, continuously remembers the states of that model, and recalls the remembered states will be conscious. There will be remaining unanswered questions, but consciousness will no longer be talked about as “the hard problem.” It won’t even be considered a problem.

Intelligence is not something that can be programmed in software or specified as a list of rules and facts. We can endow a machine with the ability to learn a model of the world, but the knowledge that makes up that model has to be learned, and learning takes time. As I described in the previous chapter, although we can make intelligent machines that run a million times faster than a biological brain, they cannot acquire new knowledge a million times faster. Acquiring new knowledge and skills takes time regardless of how fast or big a brain might be. In some domains, such as mathematics, an intelligent machine could learn much faster than a human. In most fields, however, the speed of learning is limited by the need to physically interact with the world. Therefore, there can’t be an explosion of intelligence where machines suddenly know much more than we do.

Your brain is in a box, the skull. There are no sensors in the brain itself, so the neurons that make up your brain are sitting in the dark, isolated from the world outside. The only way your brain knows anything about reality is through the sensory nerve fibers that enter the skull. The nerve fibers coming from the eyes, ears, and skin look the same, and the spikes that travel along them are identical. There is no light or sound entering the skull, only electrical spikes. The brain also sends nerve fibers to the muscles, which move the body and its sensors and thereby change what part of the world the brain is sensing. By repeatedly sensing and moving, sensing and moving, your brain learns a model of the world outside the skull. Notice again that there is no light, touch, or sound entering the brain. None of the perceptions that make up our mental experiences—from the fuzziness of a pet, to the sigh of a friend, to the colors of fall leaves—come through the sensory nerves. The nerves only send spikes. And since we do not perceive spikes, everything we do perceive must be fabricated in the brain. Even the most basic feelings of light, sound, and touch are creations of the brain; they only exist in its model of the world.

I know it doesn’t feel as if we are living in a simulation. It feels as if we are looking directly at the world, touching it, smelling it, and feeling it. For example, it is common to think the eyes are like a camera. The brain receives a picture from the eyes, and that picture is what we see. Although it is natural to think this way, it isn’t true.

I see the current human situation as a battle between two powerful forces. In one corner, we have genes and evolution, which have dominated life for billions of years. Genes don’t care about the survival of individuals. They don’t care about the survival of our society. Most don’t even care if our species goes extinct, because genes typically exist in multiple species. Genes only care about making copies of themselves. Of course, genes are just molecules and don’t “care” about anything. But it useful to refer to them in anthropomorphic terms. In the other corner, competing with our genes, is our newly emerged intelligence. The mental “I” that exists in our brains wants to break free from its genetic servitude, to no longer be held captive by the Darwinian processes that got us all here. We, as intelligent individuals, want to live forever and to preserve our society. We want to escape from the evolutionary forces that created us.

Now imagine a slightly different scenario. Let’s say we have the technology to read out your biological brain without affecting it. Now when you flip the switch, your brain is copied to a computer, but you feel nothing. Moments later, the computer says, “Yay! I’m alive.” But you, the biological-you, are still here too. There are two of you now, one in a biological body and one in a computer body. The computer-you says, “Now that I’m uploaded, I don’t need my old body, so please dispose of it.” The biological-you says, “Wait a second. I’m still here, I don’t feel any different, and I don’t want to die.” What should we do about this? Maybe the solution to this dilemma is just to let the biological-you live out the rest of its life and die of natural causes. That seems fair. However, until that happens, there are two of you. The biological-you and the computer-you have different experiences. So, as time progresses, they grow apart and develop into different people. For example, the biological-you and the computer-you might develop different moral and political positions. Biological-you might regret creating computer-you. Computer-you might dislike having some old bio person claiming to be it. Making matters worse, there would likely be pressure to upload your brain early in your life. For example, imagine that the intellectual health of computer-you depends on the intellectual health of biological-you at the time of upload. Therefore, to maximize the quality of life of your i...

Uploading your brain at first sounds like a great idea. Who wouldn’t want to live forever? But making a copy of ourselves by uploading our brain into a computer will not achieve immortality any more than having children will. Copying yourself is a fork in the road, not an extension of it. Two sentient beings continue after the fork, not one. Once you realize this, then the appeal of uploading your brain begins to fade.

Environmentalism is not about preserving nature, but about the choices we make. As a rule, environmentalists make choices that benefit future humans. We try to slow down changes to the things we like, such as wilderness areas, to increase the chances that our descendants can also enjoy these things. There are other people who would choose to turn wilderness areas into strip mines so that they can benefit today, more of an old-brain choice. The universe doesn’t care which option we choose. It is our choice whether we help future humans or present humans.