27th August 2021
The different architectures of brains and computers
The different architectures of brains and computers
What do we mean by the architecture of a system? An architecture is a way to divide up the system into parts in order to make it easier to understand how the system works. Systems can often be divided up into parts in different ways, to support understanding of different aspects of system behaviour. For example, two of the important ways in which we divide up computers into parts are the functional architecture and the system architecture.
THE FUNCTIONAL ARCHITECTURE OF A COMPUTER
In the functional architecture, the different parts are the different applications. Such applications include Safari and Firefox for browsing the internet, Word and Pages for preparing text documents; ScreenFlow and iMovie for creating videos. There are applications for playing different games, for organizing photos, or for handling finances and so on. A key feature of the way these applications are organized in a computer is that all the interactions between them occur via the operating system of the computer. For example, two applications cannot exchange information directly. To pass information, one application must provide it to the operating system, and the other application can then get it from the operating system. Furthermore, to carry out their features, applications need access to memory, processing and other resources of the computer. Such access can only be obtained via the operating system.
THE FUNCTIONAL ARCHITECTURE OF A COMPUTER
In the functional architecture, the different parts are the different applications. Such applications include Safari and Firefox for browsing the internet, Word and Pages for preparing text documents; ScreenFlow and iMovie for creating videos. There are applications for playing different games, for organizing photos, or for handling finances and so on. A key feature of the way these applications are organized in a computer is that all the interactions between them occur via the operating system of the computer. For example, two applications cannot exchange information directly. To pass information, one application must provide it to the operating system, and the other application can then get it from the operating system. Furthermore, to carry out their features, applications need access to memory, processing and other resources of the computer. Such access can only be obtained via the operating system.
THE SYSTEM ARCHITECTURE OF A COMPUTER
If you look inside a computer, a very different architecture can be seen. There are a number of pieces of flat green plastic covered with components like integrated circuits. These are called printed circuit assemblies, and they are all connected together. These printed circuit assemblies are the modules that make up the physical or system architecture of the computer.
The primary factor that determines the system architecture is the need to use physical information processing resources economically. The most complex computing systems can have hundreds of billions or even trillions of transistors, but an inefficient architecture would require vastly more. Any application uses many different information processes, and if every process was carried out by its own separate transistors the number of transistors required would be completely impractical. However, there are some similarities in the different processes used by an application, and even between the information processes used by different applications. Groups of similar processes are identified, and physical resources designed to perform all the processes in the group very efficiently. Modules are the physical resources that perform different groups of similar information processes.
If you look inside a computer, a very different architecture can be seen. There are a number of pieces of flat green plastic covered with components like integrated circuits. These are called printed circuit assemblies, and they are all connected together. These printed circuit assemblies are the modules that make up the physical or system architecture of the computer.
The primary factor that determines the system architecture is the need to use physical information processing resources economically. The most complex computing systems can have hundreds of billions or even trillions of transistors, but an inefficient architecture would require vastly more. Any application uses many different information processes, and if every process was carried out by its own separate transistors the number of transistors required would be completely impractical. However, there are some similarities in the different processes used by an application, and even between the information processes used by different applications. Groups of similar processes are identified, and physical resources designed to perform all the processes in the group very efficiently. Modules are the physical resources that perform different groups of similar information processes.
Hence each module carries out a different type of information process. A CPU module carries out logic and arithmetic processes. A RAM module carries out data reading and writing processes in a fast access memory. A display driver module carries out processes that control the pixels displayed on the screen. A WiFi module carries out processes to send and obtain data from local area network and so on. All these modules are connected together by a data link called a bus.
In computers, there are just two general types of information process: the instruction and the data read/write. Every process carried out by a computer in any physical module is some combination or sequence of these two types. The CPU specializes in relatively pure instructions, while the RAM specializes in relatively pure data read/writes.
To ensure that each module can operate relatively independently, information processes are divided up between modules in such a way that the information exchange between modules needed for them to operate is made as small as possible.
RELATIONSHIP BETWEEN FUNCTIONAL AND SYSTEM ARCHITECTURES
Any application requires information processes performed by most modules, and any module performs information processes in support of most applications. There are therefore no correspondences between applications and modules. In other words, there are no modules that carry out just one application and no others. There are no applications that use processes by just one module and no others.
Hence there is no simple relationship between the functional and system architectures. They are ways to understand very different aspects of the system. The functional architecture is useful for understanding how to use a computer. The system architecture is useful for understanding how to design a computer.
In computers, there are just two general types of information process: the instruction and the data read/write. Every process carried out by a computer in any physical module is some combination or sequence of these two types. The CPU specializes in relatively pure instructions, while the RAM specializes in relatively pure data read/writes.
To ensure that each module can operate relatively independently, information processes are divided up between modules in such a way that the information exchange between modules needed for them to operate is made as small as possible.
RELATIONSHIP BETWEEN FUNCTIONAL AND SYSTEM ARCHITECTURES
Any application requires information processes performed by most modules, and any module performs information processes in support of most applications. There are therefore no correspondences between applications and modules. In other words, there are no modules that carry out just one application and no others. There are no applications that use processes by just one module and no others.
Hence there is no simple relationship between the functional and system architectures. They are ways to understand very different aspects of the system. The functional architecture is useful for understanding how to use a computer. The system architecture is useful for understanding how to design a computer.
THE FUNCTIONAL ARCHITECTURE OF THE BRAIN
Psychology has put a lot of effort into coming up with what are often called cognitive architectures. These cognitive architectures identify major features of human cognition, and the interactions between them. A typical example identifies five different types of memory: memory for facts and word meanings called semantic; memory for individual events, called episodic; memory for skills, called procedural; short term subconscious memory, called priming; and working memory which is the ability to retain information active in the brain for short periods. Other major features include perception, or the ability to derive behaviourally useful information from raw sensory inputs, and motor control.
THE SYSTEM ARCHITECTURE OF THE BRAIN
Inside the brain, a number of major anatomical structures can be observed, including the cortex, hippocampal system, thalamus, basal ganglia, basal forebrain, amygdala, hypothalamus and cerebellum. These structures make up the major modules of the brain, and are all subdivided into submodules, sub-submodules and so on.
The primary factor that determines this architecture is the need to use physical information processing structures like neurons efficiently. The human brain has about 86 billion neurons, but if every cognitive feature used separate groups of neurons, the required resources would be vastly larger.
There are some similarities in the different processes used by one cognitive feature, and even between the information processes used by different features. Natural selection has resulted in different physical resources optimized to perform different groups of similar information processes very efficiently. Different anatomical structures are the physical resources that perform these different groups.
In the brain there are only two general types of information processes: the condition define/detect and the behavioural recommendation define/integrate. These are analogous with but qualitatively different from the process types in computers.
Information conditions are circumstances that the brain notices as occurring relatively frequently in the course of experience. These conditions are complex combinations of sensory inputs and information about the internal state of the brain itself. The brain defines these conditions only on the basis of frequent occurrence, and they therefore do not correspond exactly with behaviourally significant circumstances such as cognitive categories. In each situation, any previously defined conditions that are present are detected. Each such detection is interpreted as a range of recommendations in favour of different behaviours, each recommendation having its own weight. The brain determines and implements the behaviour with the largest total recommendation weight across all currently detected conditions. If only a few conditions are being detected, the range of recommendations is too limited for a high integrity behaviour selection. In these (generally novel) situations, some condition definitions are expanded so that they are also detected. This is the primary way in which conditions are defined.
Psychology has put a lot of effort into coming up with what are often called cognitive architectures. These cognitive architectures identify major features of human cognition, and the interactions between them. A typical example identifies five different types of memory: memory for facts and word meanings called semantic; memory for individual events, called episodic; memory for skills, called procedural; short term subconscious memory, called priming; and working memory which is the ability to retain information active in the brain for short periods. Other major features include perception, or the ability to derive behaviourally useful information from raw sensory inputs, and motor control.
THE SYSTEM ARCHITECTURE OF THE BRAIN
Inside the brain, a number of major anatomical structures can be observed, including the cortex, hippocampal system, thalamus, basal ganglia, basal forebrain, amygdala, hypothalamus and cerebellum. These structures make up the major modules of the brain, and are all subdivided into submodules, sub-submodules and so on.
The primary factor that determines this architecture is the need to use physical information processing structures like neurons efficiently. The human brain has about 86 billion neurons, but if every cognitive feature used separate groups of neurons, the required resources would be vastly larger.
There are some similarities in the different processes used by one cognitive feature, and even between the information processes used by different features. Natural selection has resulted in different physical resources optimized to perform different groups of similar information processes very efficiently. Different anatomical structures are the physical resources that perform these different groups.
In the brain there are only two general types of information processes: the condition define/detect and the behavioural recommendation define/integrate. These are analogous with but qualitatively different from the process types in computers.
Information conditions are circumstances that the brain notices as occurring relatively frequently in the course of experience. These conditions are complex combinations of sensory inputs and information about the internal state of the brain itself. The brain defines these conditions only on the basis of frequent occurrence, and they therefore do not correspond exactly with behaviourally significant circumstances such as cognitive categories. In each situation, any previously defined conditions that are present are detected. Each such detection is interpreted as a range of recommendations in favour of different behaviours, each recommendation having its own weight. The brain determines and implements the behaviour with the largest total recommendation weight across all currently detected conditions. If only a few conditions are being detected, the range of recommendations is too limited for a high integrity behaviour selection. In these (generally novel) situations, some condition definitions are expanded so that they are also detected. This is the primary way in which conditions are defined.
The cortex defines and detects conditions, and the basal ganglia interprets each current condition detection as a range of behavioural recommendations and determines the most strongly recommended behaviour. There are three general types of behaviour. One type is release of condition detections into or out of the cortex, or between different cortical submodules (called areas). Releases into the cortex are from the senses, and correspond with attention behaviours. Releases out of the cortex drive motor behaviours. Releases within the cortex correspond with steps in thinking. All these releases are carried out by the thalamus under the direction of the basal ganglia. The second general type of behaviour is reward behaviours. These behaviours change the recommendation weights of recently detected conditions in favour of recently selected behaviours. Reward behaviours are recommended to the basal ganglia by cortical condition detections but implemented on the basal ganglia itself.
Conditions are defined heuristically and therefore change over time. The problem with changing a condition is that it can reduce the validity of all the recommendation weights associated with the condition. Hence there must be careful management of when conditions will be changed and which conditions change. The third general type of behaviour is making such changes. Cortical areas associated with the hippocampus recommend changes, and the hippocampus proper determines and implements the most strongly recommended changes.
Some condition detections recommend a general type of behaviour such as aggressive or fearful types. The amygdala and hypothalamus define and detect conditions effective for these recommendation types. The most strongly recommended general types are determined by a combination of the amygdala and basal ganglia, and implemented by a combination of the amygdala and hypothalamus.
For condition change behaviours, the exact timing of the release of signals driving the changes is critical, and this timing is managed by the basal forebrain.
Many important and often used behaviours are in fact sequences of more detailed behaviours that always take place in the same order. Walking involves sequences of foot, leg, arm and body muscle movements; speaking common words or phrases involves sequences of muscle movements in the mouth, tongue, lips and throat. When such a sequence is often used, control is transferred to the cerebellum. The sequence as a whole is then selected once by the basal ganglia, and carried out by the cerebellum. Avoiding the need for the basal ganglia to select each individual behaviour means that behaviours under cerebellar control are performed much faster. However, although the cerebellum can adjust the timing of the behaviours in a sequence, any changes to the behaviours or their order requires control to revert to the basal ganglia.
The major anatomical modules of the brain thus divide up the information processes needed to support higher cognition in such a way that the total physiological resources are minimized as far as possible. In addition, to ensure that each module can operate relatively independently, information processes are divided up between modules in such a way that the information exchange between modules needed for them to operate is made as small as possible.
RELATIONSHIP BETWEEN FUNCTIONAL AND SYSTEM ARCHITECTURES
Any cognitive process requires information processes performed by most or all the modules, and any module supports most or all cognitive processes. Submodules of the primary modules are defined that specialize in information processes supporting those performed by the primary module. Hence submodules will also support many or all cognitive processes. Any search for an anatomical module or submodule that corresponds with a cognitive feature in the sense that all the information processing by the module supports only the one feature is very unlikely to be successful.
THE ARCHITECTURES OF THE BRAIN
The most useful architecture for understanding how anatomy and physiology support cognition is the physical (or system) architecture. The cognitive (or functional) architecture is useful for describing how cognitive processes operate, but is of minimal value for understanding the underlying anatomical and physiological processes. This is analogous with the situation in computing systems, where the user manual (or functional architecture) is of little value for understanding the system design.
Conditions are defined heuristically and therefore change over time. The problem with changing a condition is that it can reduce the validity of all the recommendation weights associated with the condition. Hence there must be careful management of when conditions will be changed and which conditions change. The third general type of behaviour is making such changes. Cortical areas associated with the hippocampus recommend changes, and the hippocampus proper determines and implements the most strongly recommended changes.
Some condition detections recommend a general type of behaviour such as aggressive or fearful types. The amygdala and hypothalamus define and detect conditions effective for these recommendation types. The most strongly recommended general types are determined by a combination of the amygdala and basal ganglia, and implemented by a combination of the amygdala and hypothalamus.
For condition change behaviours, the exact timing of the release of signals driving the changes is critical, and this timing is managed by the basal forebrain.
Many important and often used behaviours are in fact sequences of more detailed behaviours that always take place in the same order. Walking involves sequences of foot, leg, arm and body muscle movements; speaking common words or phrases involves sequences of muscle movements in the mouth, tongue, lips and throat. When such a sequence is often used, control is transferred to the cerebellum. The sequence as a whole is then selected once by the basal ganglia, and carried out by the cerebellum. Avoiding the need for the basal ganglia to select each individual behaviour means that behaviours under cerebellar control are performed much faster. However, although the cerebellum can adjust the timing of the behaviours in a sequence, any changes to the behaviours or their order requires control to revert to the basal ganglia.
The major anatomical modules of the brain thus divide up the information processes needed to support higher cognition in such a way that the total physiological resources are minimized as far as possible. In addition, to ensure that each module can operate relatively independently, information processes are divided up between modules in such a way that the information exchange between modules needed for them to operate is made as small as possible.
RELATIONSHIP BETWEEN FUNCTIONAL AND SYSTEM ARCHITECTURES
Any cognitive process requires information processes performed by most or all the modules, and any module supports most or all cognitive processes. Submodules of the primary modules are defined that specialize in information processes supporting those performed by the primary module. Hence submodules will also support many or all cognitive processes. Any search for an anatomical module or submodule that corresponds with a cognitive feature in the sense that all the information processing by the module supports only the one feature is very unlikely to be successful.
THE ARCHITECTURES OF THE BRAIN
The most useful architecture for understanding how anatomy and physiology support cognition is the physical (or system) architecture. The cognitive (or functional) architecture is useful for describing how cognitive processes operate, but is of minimal value for understanding the underlying anatomical and physiological processes. This is analogous with the situation in computing systems, where the user manual (or functional architecture) is of little value for understanding the system design.