Cognitive Science and Computing

The mammalian brain, when considered in the terms of digital logic information processing, can be considered as a combined memory-processor. This idea should be fairly clear if you consider the ideas outlined in my overview of cognitive 'hardware'. This is because the physical interconnections between neurons, as well as the density and type of receptor sites for neurotransmitters between them, store information in the brain long-term, whilst the activation of those same neural networks actively process information. Information storage in memory, and the activation of that memory utilise the same physical structures.

Contemporary computing architecture follows a somewhat different design. Information is stored long-term in a hard-drive. Information is loaded from memory by calls from the programming, where it is then parsed through the serial bus, to be stored into faster-access Random Access Memory (RAM) and/or parsed through a Central Processing Unit (CPU). The CPU holds within in it a number of logic circuits that are able to perform a small number of functions; like addition, AND, OR, and NOT. These functions are not all that dissimilar to the manner in which neural networks interact with each other. The major difference being, that each neural network holds both the logic and the storage, such that information processing is parallel and distributed throughout the entire brain, whereas in digital logic these things are separated into discrete units.

It is of note that human reasoning and understanding, and the products of it tend to mirror each other. It is arguable that contemporary understandings of human cognition emerged from the ideas formulated and created for digital logic and information processing. For example, the popularly held theory today amongst cognitive scientists is that there is a 'Central Executive' component to the human mind, that is very similar to the CPU within a computer. The Executive performs a number of functions onto working memory, namely; focusing attention, dividing attention, shifting attention, and linking Working Memory to Long-Term Memory (i.e., allowing active processing to make memory retrieval calls from long-term memory). Perhaps arguably as well, the logic underlying digital logic and computing was inspired by the processing architecture of the human brain.

Whilst the functions of the 'Central Executive' are undoubtedly part of the phenomenon of mind within human cognition, it has become clear that these functions are not stored in some discrete and centralised component within the brain. Instead, multiple information storage and processing modalities have these effects on each other; as discussed in the overview of cognitive 'hardware'. Study of the mammalian brain and it's structural architecture supports the notion of parallel and distributed processing from semi-autonomous subsystems far better than it does looking for some centralised processing centre. Rationally speaking, the whole brain can be considered as a Central Processing Unit for the body, rather than some discrete part of it calling all the shots and making all the decisions.

People's internal view of themselves, tends to have a consolidated and centralised sense of self. If one is to argue that digital logic architecture follows the logic of cognitive processing, it is because of this conscious and volitional aspect. Neuroscience however has come to suggest that this conscious part is something of an illusion that is stitched together by a large variety of discrete subconscious and unconscious processes in the brain. The conscious, centralised self is perhaps better seen only as the secondary process that occurs to evaluate and modulate more primary and automatic processing for the purpose of behavioural control; for example, LeDoux and the emotional brain, where automatic behaviours are first initiated by the amygdala and continue unless higher-cortical processing inhibit them. You brain processes the presence of a visual stimulus that looks like a snake, and the body freezes, your attention reorients to the stimulus and further processing determines it is only a garden hose. The result being that the automatic freezing response is inhibited, and you continue to walk the way you were going.

Computing and digital logic however, has not always followed the serial-bus and CPU architecture such as is found within the x86 architecture. Old game consoles like Nintendo followed a somewhat different paradigm. The console itself was a simple input and output handling device with limited processing. Instead, the game cartridge held most of the memory and logic circuitry required for the software. That is to say, the cartridge was essentially a custom memory and processing architecture that was streamlined for the specific needs of the game's software. The result was a system that had virtually instant load times; plug in the cartridge, turn on the console, and the software is running immediately.

At the time that digital logic architecture such as x86 was designed, there were some technological limitations in regards to things such as memory storage. Large capacity storage was almost entirely limited to disc style devices, whether floppy or hard. Whilst flash style memory did exist, it was bulky and expensive. So arguably the CPU and serial bus architecture was also a means to work around the limitations of digital memory storage. Today however, such limitations do not exist. Solid State Drives now hold large volumes of information and utilise fast loading flash memory technology.

It is now within potential to remove any need for a serial-bus and CPU, and design computers that follow the combined memory-processing and parallel distributed processing architecture of the brain. There is no need to have processing unit functions (i.e., logic gates) physically separated from memory storage, and nor is it desirable to do so, as this only creates processing bottlenecks. Instead, logic circuits can be distributed across the surfaces of a Flash Memory Drive, tying processing functions directly to flash memory clusters. If a computing device such as this was to be designed, it would allow for an architecture that is much more like the custom circuit boards of a console cartridge. The stored memory itself could call/link up to locally bound logic circuits to create soft-coded digital logic processing threads throughout the combined flash-logic unit.

Such a design has many advantages. Since memory and processing are combined physically, memory calls and processing become essentially the same thing. Without the need to parse information through the bottleneck of the serial bus, or restore it in RAM, huge numbers of parallel processing threads could be maintained. The instant loading of an old console cartridge could apply for all software stored within the unit. The only bottlenecks would be at the input and output processing sectors, much like the only bottleneck in cognitive processing is at the response-coordination modality. Power consumption would also be more efficient, modulated by the amount of active memory-processing threads within the system. Not to mention that such an architecture would allow for the generation of AI that is potentially sentient; if the software architecture followed a similar modality based approach as is found within cognition.