Could the Human Memory Prospect Artificial General Intelligence?

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.

Is an Artificial General Intelligence possible?

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. It means that whatever is being learned has to subsume into the thick set, so the time it takes to learn can be said to be correlated with the time it takes to get onto [or board] the thick set.

Aside from thick sets, there are relays with their own attributes, it is the attributes of relays that ensure they have the speed, dimensions, and so forth to properly use the thick sets and overlays.

Relays are really important, such that those for memory, particularly to produce intelligence have special characteristics, ensuring human have superintelligence. Simply, for human intelligence, memory is structured as collectives, then relays have dimensions that use collectives effectively.

AGI

It is possible to argue that deep learning is mostly relays, where identification of patterns, as well as possibility for prediction and so forth drive machine intelligence. However, whatever limitations there are, can be linked with memory.

Simply, the ways that digital memory continues to be stored as separate and single entities ensure that while training places patterns, including tying features, deep learning is doing what human memory had already done at the first stage.

So, rather than be better, they remain quite limited, drawing patterns at a lower and populated stage because classical memory is mostly the same.

Therefore, artificial general intelligence research, at least for the next year, is a battle of memory architecture, in a way that removes focus from just deep learning to storage. Basically, there can be new AGI memory research labs, to focus on this path. There can also be an early concentration on AI for oncology, developing new memory designs, for different parts, to explore if creativity or innovation germinates from AI.

Memory is also vital because it may be useful in saving energy. For example, thick sets in the human brain can be said to mitigate the necessity to process individual memory across modalities. However, for AI, memory takes a lot more, making data centers high consumers of energy.

There will be two approaches to improving memory: math and hardware. Math approaches would be to organize stored memory into collections, using Gaussian elimination, eigenvalue decomposition, duality principle, pigeonhole principle and so forth.

At the hardware level, it is possible to explore collective magnetic directions or collective electrical charges of memory cells. It is also possible to have a new VRAM equivalent for storage layers with collections. The objective is to just have data stored like human memory, and then train AI models with the collective data, and their overlays.

This is a direction to go, towards artificial general intelligence, which is likelier given how possible it is to make advances on classical memory, within this decade along with quantum memory.

There is a recent [May 27, 2026] analysis on The Verge, The Pope isn’t AGI-pilled, stating that, “On Monday, Pope Leo XIV unveiled an encyclical letter addressing the societal implications of artificial intelligence. The letter, titled Magnifica Humanitas, warned that the “use of AI is never a purely technical matter: when it enters processes that affect people’s lives, it touches on rights, opportunities, status and freedom.”

“One controversial aspect of the document, in many tech circles, was that it made no mention whatsoever of AGI or superintelligence; it allows that AI systems may “often surpass human intelligence in speed and computational capacity” but says they “lack the affective, relational and spiritual perspective through which human beings grow in wisdom.””

David Stephen currently does research in conceptual brain science with focus on the electrical and chemical configurators for how they mechanize the human mind with implications for mental health, disorders, neurotechnology, consciousness, learning, artificial intelligence and nurture. He was a visiting scholar in medical entomology at the University of Illinois at Urbana-Champaign, IL. He did computer vision research at Rovira i Virgili University, Tarragona.

See more breaking stories here.