specific motor control operation, sensory data processing, data analysis, and information interpretation. Applications show a complete coupling realization of the “hardware” and the “software.” They form cyber-physical capabilities in which the neural system will operate. A comprehensive vision system is an example where location, color, and intensity data are received into the sensors, information is transported to the brain, a neural network performs signal interpretation, a neural network performs object detection, and a neural network generates a computational response. Applications can vary between biological system drivers and artificial neuromorphic system drivers. Biological system drivers capture the functionality in human and animal brain and nervous system activity. Neuroscience is actively exploring this front. Artificial system drivers seek bounded capability targeting areas like autonomous vehicles, robotics, embedded systems, perception engineering, and image-processing computational engines. Applications are based on integration or augmentation with human sensing methods. Considering the biological neuron and neural network linkage of full stack layers, Fig. 12 shows the use case relationships between the layers, and Fig. 13 shows the types of activities and linkages between the layers. Many interdependent elements must function together to accomplish the desired computing and processing goals. Although many details are presented, it is still notional, trying to capture how a use case scenario for critical applications connects to all the lower levels in the full compute state. 8. Human brain operation layer above all other neural networks An additional layer is proposed to account for the human brain’s unique characteristics adequately. This attempts to account for human beings made in the image of God and the fact that man has a body and soul that intersect in the physical world but includes a transcendent element that extends beyond the realm of the spirit. It is essential to consider how mankind transcends the mental capability of any other created organism. Being made in God’s image, human brain function must consider the spiritual dimension. Neurons and neural network descriptions do not adequately cover this dimension. Schaeffer notes that just realizing a neural network does not mean that learning will occur all on its own (Schaeffer 2022). No human-level emergent properties are manifested by just building neural network hardware. Schaeffer may be giving evidence for brain activity that engages with non-physical input. Humans must be animated with a spirit to be alive in the sense that God intended for humanity. Leveraging theological knowledge, conceiving abductive arguments that capture how human consciousness is more than neural networks is straightforward. Although not explained in current neuroscience research, knowing that man is created in the image of God, there must be interfaces such as (1) between the human brain biology and the soul (along with the soul to spirit if considered separate), and (2) between the spirit and God. This is a possible extrapolation for what Schaeffer is pointing out when he argues that there is no free lunch. Creating elaborate neural networks is not enough to capture what is required to capture the type of learning that can be done in the human brain. Only neural networks with external interfacing capabilities allowing for abstract thinking can capture the full design of human consciousness. Fig. 14 shows a use-case scenario for the Imago Dei layer that must be included. C. Neuron models By capturing in architectural models of neuron function, it creates the first of two parts that are necessary to capture brain function from a top-level point of view. A great deal of engineering forethought has gone into its elegant and efficient operation and design. Mapping this information enables answering the first part of our third research question, what observations about human brain function can be made from the neuron and neural network architecture models? This section explores the nature of the biological neuron and characterizes its functional components using a model-based systems engineering tool that utilizes the SysML. 1. What is a neuron? A neuron is a specialized cell that receives, processes, and transmits nerve impulses. In a learning mode, it develops connections with other neurons to collaborate in computation efforts as needed (Davies 2006). From a computing point of view, a neuron can be considered a computational device that consists of a processing unit, memory, and input and output devices. In contrast to a centralized Von Neumann computing architecture with separate processing, memory, and input and out modules, each neuron is a self-contained computing agent that can be networked with other neurons to form networks. How a neuron implements each of the three Von Neumann architecture elements is explained next to clarify what is meant by comparing an individual neuron with a computational element. Stallings is used as an in-depth reference for electronic computer organization and architecture that explores the implementation and refinements that have taken place over many decades (Stallings 2019). First, neurons process and condition the information it receives. This UC Neuromorphic Computing Full Stack Top-Level Activity Human Body 1. Use Biological System Materials 2. Task Components and Devices 3. Task Circuits 4. Task Microarchitectures 5. Task System Architecture 6. Utilize Algorithms 7. Task Applications Human Being Figure 12. Neuromorphic computing use case with full compute stack. JOHANSEN Human brain function and the creation model 2023 ICC 299
RkJQdWJsaXNoZXIy MTM4ODY=