Page Nav


Classic Header




The integrative neurons and Al's deep learning evolution

 Experiments show that there are neurons in the brain that are active only when a certain situation is present. For example, there is a neur...

 Experiments show that there are neurons in the brain that are active only when a certain situation is present. For example, there is a neuron in the brain that is active only when we think about, see, feel, sense a certain object in some form or way. Or a specific person. Or a specific anything. It looks like everything in the world that we know has a specific neuron in our brain. It is active only when that specific thing appears, is perceived in some way outside, or is represented in some way inside. The representation somehow description means a real anyway, so not only when we actually experience the given object, but even when we just think about it without the actual presence of that object. 

It can be said that there is at least one specific neuron in our brain for everything that can be generalized, classified, categorized by the brain. That specific neuron is active only when that specific object is present in any representation. The desk neuron is only active when we see a desk of any kind, size, color, or direction, or when we just think about a desk. Also, the brain has a special Alice neuron (Alice must be a specific person), and if the person who has that Alice neuron sees her, hears her, or just reads about Alice, or even just thinks about Alice, so that Alice is represented in some way, that Alice neuron becomes active. 

This setting also works the other way around. When a particular integrative neuron is active, its activity can activate the corresponding neural circuits, and the active integrative neuron is able to bring up the corresponding brain representations, memories about that particular generalization.

The advantage of having these integrative neurons is defining. It allows us to think abstractly, to possess meaning, and it also allows us to have abstract language. These are important aspects of having a high level of intelligence. It is difficult to overstate the importance of the existence of integrative neurons. How can they exist? How are they created? 

They must not exist primarily in a preset way. It is difficult to preset the brain to have the integrative neurons without experiencing the explicit subject. Evolution would be able to create such presets, and evolutionarily it would be advantageous to have some integrative neurons at birth. These evolutionary presets would precondition the organism for certain abilities, such as recognizing certain dangerous situations from the time of birth. 

However, the integrative neurons must also be created dynamically in the context of learning. With this ability, the organism is able to generalize, create abstractions, and have the ability to use these generalizations and abstractions dynamically. 

The function is important, but the process of creating these neurons should be simple, should be a basic procedure based on the method of how neurons work in a network. The creation of integrative neurons should be an emergent phenomenon of the complex structure of the brain and the simple function of neurons. This property could be the basis of abstract thinking and abstract language of a sufficiently complex brain. Even if the integrative neurons can be present in less complex brains, more complexity gives higher level abstractions. 

Artificial intelligence, especially deep learning, has the same goal. It is about creating generalizations and correlations from diverse inputs. The deep learning method is modeled by the brain, but the actual implementation of the task is based on mathematics, mainly on sequential probability calculations on modeled artificial neural networks, practically using specific digital computer architecture. The method works, but it has its own limitations. It requires many calculations that are difficult and slow to perform on digital computer architecture. This is the main bottleneck of the deep learning computational method. The results of Al's deep learning method are remarkable, but the brain surely works in a different way. 

How does the brain create integrative neurons, as it is a fundamental step in creating classifications and contexts, creating meaning? What is the emergent process of the cooperating neurons, what makes the brain capable of creating the integrative neurons?

The way it works must come from the basic principles of the mechanisms of the neuron functions, the way they are connected, and the way they work together to form a network. It has to be an emergent property of the settings. 

We know that neurons that fire together tend to form connections with each other. We know that neuronal connections have one-way directions. We know that connected neurons can stimulate or inhibit the activity of other neurons. We know that neurons operate in an all-or-nothing fashion, firing only when a threshold is reached by the activating and inhibiting inputs. And finally, we know that the neurons are constantly working, firing frequently, the actual inputs only modifying the frequency of firing. These neurons form a structure created by evolution. Based on this structure, based on the number of neurons and their actual ability to make new connections, the functioning neurons are constantly modifying the structure and giving plasticity to the interconnected network, to the brain. We know that more plasticity provides more adaptability and more intelligence to the brain and to the organism. 

However, we still do not know what is the main principle of how the brain organizes itself into a plastic network of an intelligent system of cooperating neurons. The principle must be an emergent property of the complex cooperation of the basic mechanisms of neurons. 

This mechanism must be the resonance. It has been discussed in the thoughts that the resonance model can be the method of the functioning brain, and the model could explain the missing connections between the simply functioning neurons and the intelligently functioning brain. This emergent mechanism, the resonance as the main organizing principle, could also create the integrative neurons effectively, and can provide the effective functioning of the efficiently working, naturally existing architecture of the brain.

Deep learning artificial intelligence has mathematical methods to be capable of creating similar functions. The effectiveness of these methods has significant constraints compared to the brain, but having significantly stronger computational force, the potential is present. By forming integrative functions subsequently of the practiced methods of finding correlations, artificial intelligence would be capable of forming the required classifications to possess the property of meanings. The function of meaning would be a defining property of the development of artificial intelligence.

No comments