Small Islands of Consciousness

Memory, and other awareness-driven learning

From Narrow To General AI
6 min readJun 3, 2023

This is the fifth post in a series on AGI. Read the previous post here. See a list of all posts here.

Summary of posts 1 to 4: To tackle the mountain of challenges related to developing Artificial General Intelligence (AGI), begin with a set of mental functions that are invariably present in every mind. These are usually invisible, automatic mechanisms which drive the mind forward. Analyzing the mind in granular detail requires stepping back from complex, abstract structures like “concepts”, “plans”, or “memories” and focusing on individual thought-actions and their interconnections. These can give insight into some hitherto invisible aspects of the mind; for example, that the mind requires at least two independent learning “components” for introspection to be possible.

It is obvious that the mind changes over time. We generally call this “learning”. What exactly happens when you learn, and if there are more than one type of learning is still an open question. The focus of this series has been on the invisible, basic, and above all automatic mechanisms that drive the mind. So let’s start by looking at a simple, atomic type of learning, that is usually referred to as “becoming aware of something”¹.

Becoming aware of something happens in a single moment in time. Prior events may prepare the mind for becoming aware in a given moment; but the moment itself happens in one shot. Memories are the best-known type of awareness-based learning. For example, you may remember the sight of a car crash, or that time you went on an amusement park ride with your parents.

This points to the first noteworthy feature of awareness, which is that it has “content”. During learning, the mind changes in a way that is distinctly related to its cause; i.e. the “something” that you are aware of. This may originate from the outside world (e.g. noticing a friend’s new hair colour); or you may become aware of something inside your mind itself (e.g. realizing that you were daydreaming).

In all cases of awareness, you only know that a change happened because you are later able to observe some alteration in yourself. You may discover this through introspection e.g. you see a car go round a corner, and later you are able to think about the colour of the car. Or, more indirectly, you may see it change your actions e.g. you are introduced to a person by name, and later say their name to others.

So the steps involved in an instance of awareness are: (1) you experience something either in the world or in your head, (2) some invisible process that you can’t control “records” it, then (3) the next time you look inside your mind, or when you are in a similar context, the new pattern of experiences can be observed which was not present there before. For example, you may become aware of a bird on a building, then, when you are in front of a similar building, you imagine that bird in your mind. Or you might mull over a long-forgotten song while waiting in line, then while waiting in another line, you might recollect that song.

You may have noticed something odd in the two examples above. If memories are distinguished by their concrete content, then both the examples given qualify as memories, even though the source of the latter (the song) was internal. This is not the usual interpretation of the term “memory”. Yet as you saw it fits the criteria as neatly as when applied to inputs from the external senses. It seems there is something more broadly significant about the process of awareness than we originally assumed. This insight is worth digging into.

When someone refers to introspection as “seeing” into the mind, this is generally a metaphor, since there is no literal “inner eye”. The metaphor only seems appropriate because, as mentioned above, awareness of internal experiences seems to resemble the process of learning about the outside world. Every act of seeing something in your own mind creates a new “learning” or change. When you observe a memory, you now know that you have a memory; as in, you just learned something new about your mind, and that may change your thoughts going forward. Or you may note something particular in the memory, such as that the car in it was red, a fact that was hitherto only implicit³.

Since, in such cases, the content you become aware of is generally the same as what was already there in the memory, this process is usually called “recollection”, and not awareness. But that is misleading, since it makes you think that no change has happened during the process of remembering, when it clearly has; e.g. when you struggle to remember a name and it suddenly comes to you.

To remove any doubts about this — namely, that you can form memories of experiences inside your mind — let’s look at one more curious phenomenon: remembering your dreams. You can remember (i.e. become aware of) events in your dreams just as you can of waking life. This is part of what makes dreams seem so lifelike at the time, despite how wildly they diverge from waking life in terms of consistency, logic, plausibility, continuity with your waking memories, etc. The ability to be aware of events in dreams is also the reason you still feel like you have “consciousness” while dreaming, despite all the nonsense and chaos in them.

The connection between awareness and consciousness has long been recognized. Though they don’t perfectly overlap, it seems impossible to imagine being conscious without being aware of something — anything at all, be it your internal state, an idea, the external world. However, this confronts us with a strange inconsistency. Awareness, as described above, is restricted to short, self-contained segments of learning: you see/hear something (or think something), your mind presently learns about it, and you can recall it later on. On the other hand, the lay perception of consciousness is that it is a continuous state of being.

Perhaps this second belief is misinformed. It seems that although you can be conscious at any time, it is not true that you are conscious at all times. If awareness is a momentary phenomenon, then consciousness is not a continuous state of human experience. Rather you are aware in small chunks, like “islands” of awareness surrounded by a sea of unknown.

This post has described how your mind can change through moments of awareness. By itself, this is not yet sufficient for a full model of human learning. There are still some challenges to be addressed. First, although awareness may be a useful pattern to explain how you learn concrete memories, like that of a building, it still falls short of explaining how you can be aware of more insubstantial things like feelings, concepts, qualia, and perhaps even consciousness itself.

There is also the unanswered question of what exactly is happening during this process. The changes it creates in your mind, although related to the source, are in no way a perfect duplicate of it. If you see a person yell, the memory in your mind may be only one static frame of that encounter. Or it may be embellished with additional details, like a thought of the cause, or the word “angry”, or other interpretations. You should also be suspicious about the process of introspection that later informs you what the content is.

Finally, we have yet to explain why you learn any of these things in the first place. There is clearly too much in the universe to become aware of it all, and even that which you learn appears to be mutated into an abbreviated form that is useful to you. Humans are selective in their attention — that is part of the definition of attention. So what triggers awareness— i.e. what makes you become aware?

The next two posts will dig into the second and third questions: what awareness itself is, and how it is triggered. Awareness is a critical aspect of both empirical and introspective learning, so it is worth thoroughly uncovering its details and mechanisms.

Next article: T̶e̶n̶s̶i̶o̶n̶s̶ Thinking as self-generated experiences

¹ Ultimately this series will show that becoming aware is one of only two types of learning. The other is when you learn to like and dislike things. To prove this, however, will require expounding the details of planning, calculating, contemplating, etc; so for now you can assume, if you like, that awareness is one of many types.

² This will be seriously questioned in a later post.

³ This is similar to Jung’s split between the conscious and the unconscious.

--

--

From Narrow To General AI

The road from Narrow AI to AGI presents both technical and philosophical challenges. This blog explores novel approaches and addresses longstanding questions.