If the cortical Neuron have the ability to learning and forgetting data by themselves (making and breaking synaptic connections) , then why brain do have a separate structure called hippocampus for making short and long-term memory. And why don’t HTM theory accommodate sub cortical structures like hippocampus that essential for cortical functions like learning and forgetting?
Just like a transistor on a silicon chip, a neuron by itself is not very functional. It’s in cooperation with other neurons that the complex memory processing becomes interesting.
Some neurons act together as a convertor, to code information from the senses. Other neurons act as a filter, and others still act as a kind of stabiliser. In conjuction this forms SDRs (sparse data representations) where each bit has a specific semantic value.
In certain layers of the neocortex, these SDRs can be compared in sequences to detect temporal correlations (changes over time that are related). In other layers the sensory information is processed in conjunction with location information (so-called grid and displacement cells), so the neocortex interprets features in relation to each other in space.
I am a beginner in this fascinating field, and I have still lots of questions myself. But one of the things I start to understand is that the speed at which synapses change is an important difference between short and long term memory. And apparently that’s why there are different locations and types of memory in our brain.
In a computer, we have well defined functions blocks that are islands of function. The CPU is over here, the cache memory is close by, the main memory is over there, and the rotating memory is way over there. Once you get used to thinking this way is hard to think of computing working any other way.
The brain does have different functional units but the grouping are very different from what you may be used to if you have worked with traditional stored program computing. The brain weaves the different parts together with the memory and computing units being the same. In the subcortical structures, there are some well-defined functions such as the amygdala. These are still memory and computing mixed. In the cortex, there are about one hundred islands of functional units with the biggest difference being what is connected to a given functional unit; what a given map processes is determined by what it is connected to. The connections may be to some sensory or motor part of the body, or to other maps. These processing units are called by names such as area, map, or region.
I have written two posts that may give you some orientation into how the various memory functions work together in these units.