Could the Human Memory Prospect Artificial General Intelligence? Is an Artificial General Intelligence possible? More about Irish Tech News

By David Stephen
Conceptually, humans have generalized memory. Simply, humans do not have specific memory of everything in the external world, however familiar. There are groups or collections in human memory that make it easier for humans to navigate the world. These collections also make access thorough for relays or transport, than having the memory area cluttered with respective memories of everything. This says that the function called human intelligence is foundational on memory. So, if human intelligence is generalized, then human memory is also generalized. This indicates that if artificial general intelligence [AGI] will be possible, a lot of effort has to include new memory architecture, similar to human memory. While there would be better deep learning architectures than transformers, it should be evident that the excellence of transformers, even with classical memory would have been better if there was a different memory structure.
So, no matter the promise of world models, or the promise of neurosymbolic AI, memory is so integral to intelligence that it is unlikely that both directions would archive AGI.
Intelligence can be defined as the use of memory for expected, desired or advantageous outcomes. This means that the way memory is used, determines what becomes intelligence. The better memory is used, the more intelligent or the more effective the memory is, for whatever outcome.
If memory use means intelligence, it implies that memory is a station or destination, then intelligence is transport across those destinations. It is true that in the brain, there are many destinations — so to speak — but those for memory and how they are visited make determinations for intelligence.
Now, for all there is to know about the external world, assuming that every memory is separately held, so a violet door, a green door, a long door, a wooden door, a metal door, and so on, then if intelligence would keep visiting all, it would be too slow to respond and sometimes not reach where it should.
This is a reason that human memory is stored as a collection of similarities. Simply, assuming a destination is a thick set, then a thick set collects whatever is similar between two or more thin sets.
Conceptually, a set refers to an assembly, configuration or formation of electrical and chemical signals. A set is also theorized to be obtained in a cluster of neurons. A cluster of neurons may have one or more sets. However, a set is how information is organized.
Now, because a set organizes a door, there are respective chemical signals and electrical signals that must assemble in a particular way, to result in that door. Then, if another door is seen, there would be similar assembly [of signals] even as they are not the same.
So, what the brain does is that it collects all those similarities, between any set into a bigger one. This is what becomes utilized to make interpretations. This implies that whenever any door is seen, what is used to know it is a door is the thick set. While there are several thin sets for specific things — with no similarities — they are usually fewer.
Aside from thick sets, there are overlays of thick sets, which often changes. For example, the thick set of door overlays with the thick set of knob. The thick set of door can also overlay with the thick set of metal or wood, key, or lock and so on.
Overlays are often temporary. And sometimes switches fast as well. They allow for memory to be positioned for use, more easily and to ensure that creativity or innovation is possible or even when something is routine, there is at least the chance to have it feel different.
Overlays are also useful to ease how to figure things out, or come up with something new even without the intention of doing so. Memory can overlay in certain ways, making relays pick on that to use it.
While thick sets have several advantages, one of their major disadvantages is learning, especially something new and not as familiar, for an adult. I...