computing. With so many basic features in neuroscience that still are unknown, capturing architecture models can provide a framework for how to view this complex information. Since the biological neuron implementation details are complex and not wholly characterized, architectural modeling offers a different approach to identifying design methods and interdependencies. Neurons are a fundamental building block in every part of human neural systems. In general, they accomplish the same function using a common basic architecture. There is specialized tuned functionality when moving towards more individualized neuron types as one compares capabilities like sensing, motor control, computing, and learning. Below are the insights that can be drawn from functional and structural classifications. a. Functional classification One way of showing functional neuron classifications is to bin neuron types into sensor, interneuron, and motor neurons. Interneurons have two sub-classifications, one in the central nervous system and the other in the brain. Even at this high-level view, there is a tree structure where all neurons have a variety of common characteristics. Still, then there are additional specialized features that are utilized for their specific mission. Sensor neurons cooperatively work with the sensors. Motor neurons are tailored to work with muscles and ensure proper control. Interneurons focus on relaying information and doing a portion of the computational load. The central nervous system is primarily a transport mechanism, while brain neurons take relayed information and perform computations and learning functions. b. Structural classification Structural neuron classifications focus on physical implementation differences and how they vary from one neuron class to another. Major structural differences group neurons into unipolar, bipolar, and multipolar. A unipolar neuron has a single dendrite. A bipolar neuron has one dendrite and one axon, which is useful for direct and indirect cell pathways like in the eyes. Multipolar neurons are typical in the nervous system and have long axons. This type of architectural classification approach groups neurons by their physiology. c. Thoughts on the divergence of biological versus artificial neural system goals Neurons are very resource efficient in accomplishing their purposes. Neuromorphic computing implementation of artificial neuron finds it very difficult to accomplish. Some features do not translate well when electronic neuron-like materials instead of biological neuron materials are used. In contrast to what Pavone and Plebe suggest by proposing that neuromorphic computing systems should minimize trying to achieve their goals by mimicking biological brain architectures (Pavone and Plebe 2019), it is impossible to separate the reason why God created the human brain from its implementation. Understanding its creation context makes a huge difference. As the Creation Model shows, all creation in its original context works together harmoniously with key performance goals in mind. God wanted to engage with humanity. God did not want a man to be alone (Gen 2:18). God wanted man to fulfill his Imago Dei calling (Gen 1:28). God did not intend as a central focus for man to become augmented with technology just to increase personal capability. Instead, it is all about relationships and drawing all aspects of creation back to greater intimacy with God. 4. Drawing the observations of the three research questions together Human beings were designed with mental capabilities that exceeded all other animals. They were created to live in harmony within the creation fashioned for them on Earth. Utilizing the full compute stack model, the capabilities of humans go beyond what its layers can capture. With the focus on neurons and neural networks in this paper, it is possible to consider how the Imago Dei translates to this level with an architectural model. Much about neurons and neural networks still needs to be uncovered. Creating an architectural context along with what is already determined with neuroscience research and the challenges that have occurred with neuromorphic computing trying to implement comparable systems gives insight into how much capability is packed into the human brain. It is asserted that generative artificial intelligence will never match what humans can do. There are missing architectural layers that cannot be included in artificial systems. Human responses can be codified, and aspects of learning can be captured in machine learning approaches. Still, this does not mean that artificial intelligence systems will have the breath of life from God given to them (Gen 2:7). V. CONCLUSIONS Above all other parts of creation, the human brain alone can think and process abstract ideas. Humankind should be the last part of creation since there are interdependencies between man and every other aspect of God’s handiwork. More importantly, humankind is called to rule and reign over creation and held accountable to be a good steward. Individuals report directly to God and carry the bidding of our sovereign God to the ends of the Earth. Neurons are a fundamental building block for the various parts of computational tasks. Biological neurons are a complex and adaptable building block used in many human and animal physiology places. Several categories and types within each category exist. This is a very active area within neuroscience. Artificial neurons and neural networks are not able to meet the capability and the modest resources necessary that are found in their biological counterparts. Neural networks are collections of neurons that form cooperative structures that allow them to operate together through learning and optimization from repetitive tasks. Exploring neural networks, and computation in general, in terms of full compute stack layers is a helpful way to capture the levels of functional capability that must cooperate to bring together a working computational capability. Specific engineering choices are seen in the implementations of these layers. Research continues to provide new insight into how these functions work. Although this paper integrates various ideas for framing an understanding of neurons and neural networks, the work requires more development because there is a limited understanding of neuron and neural network biology, and artificial implementations are rapidly changing. Continuing to track the developments in neuroscience, machine learning, and neuromorphic computing will be required. Follow-up survey papers and focused research question-centered papers will continue to be developed. JOHANSEN Human brain function and the creation model 2023 ICC 309
RkJQdWJsaXNoZXIy MTM4ODY=