1000 Brain Theory

One of the most difficult problems in the world of science is the problem of how human intelligence works… Isn't it ironic? The human brain still does not fully understand how the brain works. How is it that something the size of a towel is forming our thoughts?

How can I compare the brain to such a towel? What happens is that if I crumple this towel a little with both hands, a kind of structure with folds in itself emerges. Isn't this the first shape that comes to mind when we think of the brain? There is such an outer layer in the brain of mammals, and the largest one is in us humans. This outer layer, which covers 70% of the human brain, is called the "neocortex". In other words, a word meaning "new shell, new membrane". If you open this folded membrane, it looks like a towel of this size, with a thickness of approximately 2-3 mm. Everything about human intelligence takes place in the cells that are in this towel-sized area. We understand the world around us, the universe, with this towel on our head. We make new discoveries with this towel.

The neocortex, which is responsible for all cognitive processes such as science, mathematics and literature, also manages our language and perceptions. In short, this is the part of our brain that is related to intelligence. There's also the "old brain" part, which makes up the rest of it, 30%. Here, our actions such as walking, running, chewing; it manages our reflex behaviors and emotions such as breathing and digesting.

If we want to understand human intelligence, we need to understand this outer layer. There are functional regions on the neocortex, which looks monotonous, just like this towel when viewed from the outside. Some places enable us to see, some places enable us to hear or speak. The interesting thing is that when we look closely at these regions with different functions, we do not see different types of cells. Everywhere we encounter the same kind of structure: a columnar structure.

This microarchitecture is consistently present throughout the neocortex. These structures are not as visible to the naked eye, but each column does the same thing as a function. If we compare this to an electrical circuit structure, the same circuits enable us to do different jobs in different regions. There are around 150,000 cortical columns in our brain and around 100,000 neurons in each column. Each neuron makes thousands of connections with the others. These connections enable different operations to take place between the cell layers. If we can figure out what neurons and layers do, in short, how a single column with a diameter of 0.5 mm and a height of 2 mm can work, we will have taken a very important step in understanding intelligence. Because these columns exist not only in the human brain, but also in the brains of all mammals. The more the same structure is copied, the larger the cerebral cortex, and therefore the more advanced the intelligence.

A lot of research has been done on this subject for nearly 100 years and many theories have been put forward in these researches, but I can say that the last one is very ambitious. This theory, called "1000 Brains," not only claims to explain how human intelligence works, but also says it can be applied to machine intelligence. Already the owner of the theory, Jeff Hawkins is actually the inventor of the handheld Palm computers, which is considered the ancestor of mobile phones. Later, he directed his studies to the field of neuroscience and founded a research company called Numenta to reverse engineer the human brain.

The first research published by this group is about why neurons make thousands of connections. Accordingly, each pyramidal brain cell works as a kind of prediction machine. It understands hundreds of different signal patterns and makes predictions based on these inputs.

To understand the group's second major research, we will use an object. When you pick up a cup and hold it, you can imagine how you would feel before touching it with your finger, right? In order to make such an estimate, you must first know the position of the finger. This process, which seems very simple, is actually quite complex. Because location information is not absolute. Relative information based on the location and orientation of the cup and finger. In order to create this information, our brain creates a frame of reference for the cup.

According to the "1000 Brain Theory", our brain learns through action. I create a sensor input when I touch the cup. This input is transferred to the reference frame as a positional information in the cortical column. Every movement of your finger is constantly updating this information. It creates a model of that object.

When three fingers touch, three models are created, and when five fingers touch, five models are created. By seeing the cup on the one hand, hearing it on the other hand, feeling its warmth and coldness, not one, but thousands of constantly updated models are created in our brains. That's why the theory is called "1000 Brain Theory."

Touch is easy to understand, but how does this mechanism work in our senses such as hearing or seeing? According to these researchers, seeing is almost the same as touching. When we look at the cup, we do not take a single glance. Our eyes move like our fingers at least 3 times a second, focusing on different points of it and creating different models. While hearing or smelling sounds, we also collect inputs from the world around us through similar movements.

This "learning by motion" part of the theory seems very plausible to me. Because the input to our brain is constantly changing. There are two reasons for this. First, the world outside of us is constantly changing. For example, while listening to music, the inputs from our ears send us variables according to the melody and rhythm of the music, that is, its movement. Similarly, a tree swaying in the wind causes both visual and auditory changes. In both examples, the input to the brain changes from moment to moment, not because we move, but because things in the world move.

The second reason is that we move too. Every time we take a step, move a limb, move our eyes, tilt our head, or make a sound, the input from our sensors changes. As I just said, our eyes fixate on a new spot in the world about three times a second, and the information going from the eyes to the brain changes completely. If we hadn't moved our eyes, such a change would not have happened. This movement of our eyes continues even during the night, let alone during the daytime. Even while dreaming, we are doing a kind of learning and modeling process of the world.

Because the brain learns its own model of the world by observing how its inputs change over time. There is no other way to learn. We cannot upload files to the brain as we install a program on a computer. “Unlike a computer, we cannot upload files to our brain. The only way a brain can learn anything is through changes in its inputs. If the inputs to the brain were static, nothing could be learned.

Yes, our body may not move while learning the melody of a song we love. How do we learn when we sit completely still with our eyes closed? In fact, the movement continues. By listening to how sounds change over time, we learn that new melody. But that's not always the case, is it? So we are fidgeting, we can't stand still and start dancing. Because learning requires us to act and explore actively. Imagine entering a new house that you have not entered before. If you don't move, there will be no change in your sensory input and you will probably not be able to learn anything about the house. To find out a model of the house, you need to look in different directions and walk from room to room. You have to open doors, look in drawers and turn things around. The house and its contents are mostly static; they do not act on their own. That's why we need to move in order to learn the model of a house.

We now move on to an even more interesting claim of the theory. We said that, according to this theory, we create a model with every movement of our sense organs. Our sensors create thousands of models with the inputs they collect. This theory says that each different model created votes among themselves to reach a decision. Yes, you heard right. According to this theory, our brain is governed by democracy.

Because our fingers, eyes, briefly, our sensors are constantly moving and collecting input, but the information we create about the cup object is fixed. Here we reach this information by voting results of thousands of different models.

Let's simulate this now. Let's say I'm about to touch this object with my finger. As I move my finger towards that point, a position data emerges in my brain. We said that pyramidal brain cells work like a prediction machine. When a touch occurs, a sensor input gets there and is saved as feature data. But with just these two data I can't tell that this object is a cup. For example, it could be a beverage can or a tennis ball.

If I move my finger to another point of the object, I get a new location data, and when I tap there, I get a new property data. Move and touch once again and I have created the third position and feature dataset. These models, by voting among themselves, eliminate the wrong guesses about the object and reach the right decision and reach the conclusion that this object is a cup.

One way or another, they have to arrive at a right or wrong conclusion. They can't stay together. In fact, if you want, I can make such a vote take place in your brain right now. What do you see when you look at this figure? Yes, right now your brain is trying to reach a consensus. Some of those tiny pillars on the towel on your head shout “vase vase”, some shout “human faces, human faces”… They vote among themselves and let you know the result. Because the voting layer wants to reach a consensus - it doesn't allow two objects to be active at the same time - it therefore prefers one possibility over the other.

Maybe this is exactly why we look at the same things, see the same things, but reach different results. We all have the same towel on our head, and that towel is made up of thousands of columns of loops that are duplicates of each other. But voting with the data sets created by the sensor inputs to these columns gives rise to different information for all of us. We are modeling the world not with one brain, but with a thousand different brains, and we are constantly updating these models.

I guess the real issue is transforming this thousand knowledge, created with a thousand brains, into wisdom…