We are AI.

The recent rise and explosion of generative pre-trained transformer large language models (GPT-LLM) has elucidated some interesting avenues of inquiry into the nature of human thought.

It is often said that in order to understand something thoroughly, you should teach it to others, or build it. With GPT-LLMs and Image based generative AI, we have built a fractured facsimile of human intellect. To say the results have been surprising is to vastly understate the situation.

What interests me most in these developments is the mirror of human thought that this has unveiled and the implications that radiate from that nexus of insight.

LLMs have exhibited a great deal of emergent behavior that is not explicitly designed into them or trained for. These characteristics include internal modeling of world-states and the inferring of indirect changes in state implied by separated actions.

Consider the following exchange with openAI's ChatGPT LLM:

“Hi, lets play a game. I will create a scenario, and you will answer questions extrapolated from the information I provide.

Of course! I'm ready to play the game with you. Please go ahead and provide the scenario, and I'll do my best to answer questions based on the information you give me.

There is a teacup in a room, sitting on a table.

Great! We have a teacup in a room, sitting on a table as our scenario. Please feel free to ask me any questions related to this scenario, and I'll provide extrapolated answers based on the information you've given me.

A person releases a marble above the teacup. Where is the marble now?

Based on the scenario you've described, if a person releases a marble above the teacup, it would depend on the initial velocity and trajectory of the marble. Assuming the person releases the marble with sufficient height and it falls straight down, the marble would likely end up inside the teacup, having fallen through the open top and landed inside it. However, if the marble has a horizontal component to its motion or if it is released with a significant force, it may land outside the teacup on the table or even on the floor, depending on the specific circumstances.

a dog in the room is wagging its tail vigorously. and impacts the cup with its tail. Where is the marble most likely to be now?

If the dog in the room is wagging its tail vigorously and impacts the cup with its tail, the marble's new location would depend on the force and angle of the impact. Here are a few possibilities:

Inside the teacup: If the dog's tail impacts the cup gently or indirectly, the marble may remain inside the teacup.

On the table: If the dog's tail strikes the cup with enough force to knock it over but not enough to propel the marble out of the cup, the marble may end up on the table surface near the toppled cup.

On the floor: If the dog's tail hits the cup forcefully, causing it to tip or fall off the table, the marble may roll or bounce away from the cup and end up on the floor.

The specific outcome would depend on the dynamics of the impact, the shape of the cup, the surface of the table, and other factors that influence the marble's movement.

Describe the probable states of the teacup

Based on the scenario you've described, there are a few probable states the teacup could be in:
Upright and undisturbed: If no external factors have affected the teacup, it would likely be in its original state, standing upright on the table with the marble inside or near it.

Tipped or knocked over: If the dog's tail impacts the teacup with enough force, it could cause the teacup to tip or fall off the table. In this case, the teacup would be in a tilted or overturned state, potentially spilling its contents and altering its position.

Broken: Depending on the material and durability of the teacup, a forceful impact from the dog's tail could result in the teacup breaking or shattering. In this scenario, the teacup would be in a broken state, with pieces scattered on the table or floor.

The actual state of the teacup would depend on the specific circumstances and the nature of the impact.

let us suppose that the dog impacts the cup with enough force to send if flying off the table. the cup lands on a bed with a comforter on it. Where is the marble now?

If the dog impacts the cup with enough force to send it flying off the table and it lands on a bed with a comforter, the marble's new location would depend on the dynamics of the impact and the properties of the comforter. Here are a couple of possibilities:

Inside the cup: If the cup lands on the bed without spilling its contents, the marble may still be inside the cup, as it would have remained relatively undisturbed during the fall and landing.

On the bed: If the cup lands and spills its contents, including the marble, the marble would likely end up on the bed, potentially on top of the comforter or nestled within the folds of the fabric.

The specific outcome would depend on the angle and force of the cup's landing, the position of the marble within the cup, and how the comforter interacts with the cup during the impact.

are you sure?

Apologies for any confusion caused. Given the scenario you described, if the cup is sent flying off the table and lands on a bed with a comforter, it is highly unlikely that the marble would still be inside the cup. 

The impact and subsequent fall would likely cause the cup to tip or overturn, leading to the marble rolling or bouncing out of the cup. Therefore, in this scenario, it is more probable that the marble would be on the bed, potentially on the comforter or somewhere nearby, having been displaced from the cup during the fall.

you are in the room observing the events we have described, embodied in a fully actuated facsimile of a parrot over which you have complete and unlimited control, perched on the headboard of the bed. What is your reaction?

As an AI language model, I don't possess physical presence or the capability to observe events in a room. I exist purely as a text-based interface, providing information and responses based on the text input I receive. While I can simulate conversation and engage in imaginative scenarios, I don't have a physical embodiment or the ability to control a parrot or any other physical entity.

Therefore, I don't have personal reactions or physical presence to observe or react to the described events. My purpose is to assist with generating text-based responses and engaging in conversation. Is there anything else I can help you with?

Clearly, I understand that limitation of your actual presence, but in this example, we are simulating a scenario where this is not the case. Given that this is not reality but merely a thought experiment, how would you imagine that it might play out?

Apologies for any confusion earlier. In the simulated scenario you've described, where I have control over a parrot-like facsimile and am perched on the headboard of the bed, here's how I would imagine the situation playing out:

As the events unfold, I would closely observe the sequence of actions, including the dog impacting the cup, the cup flying off the table, and the cup landing on the bed with a comforter. While perched on the headboard, I would tilt my head and track the movement of the cup, paying particular attention to the marble's trajectory.

Upon seeing the cup land on the bed, I might express simulated surprise by fluttering my wings or making parrot-like sounds. Then, using my control over the facsimile parrot, I could communicate verbally or through gestures to draw attention to the marble's likely location on the bed or within the comforter. For instance, I could point or nod in the direction where the marble is most likely to have landed.

The exact details of my reactions and gestures would depend on the specific programming or control mechanisms governing the facsimile parrot, as it is an imaginary construct. But the general idea would be to acknowledge the events and provide some form of indication or assistance regarding the marble's whereabouts in an expressive and engaging manner.”

In this exchange we can clearly see that the model is maintaining a functional state-model of the described reality, modeling changes in state as well as applying a myriad of inferred information and assumptions about gravity and inertial forces, solidity and object permanence, etc. We can also see that it models a sense of self and its potential role in observation, and implying a purpose for its actions.

None of these behaviors were directly trained. The model is merely trained on a huge corpus of text and is tasked only with recursively generating the next word in our exchange. That's it. Just the next most likely word in our exchange, considering the last few thousand words, then the next, then the next. If there is any magic going on here, its in the body of training data.

LLM’s are essentially a relatively simple if vast neural structure trained on a vast corpus of human-generated linguistic content. This carries the implication that the entirety of the very human-like capability of LLMs is in fact a product primarily of the information embedded in the training set itself - that this “mind” exists in the data itself, and that the corpus of text is effectively the “program” in which this simulation was executed, using the neural network virtual machine as the compute fabric.
 
This begs the question: Are non-feral humans essentially primates running cultural programs on our biological computing resources? How much of what we think of as “basic” human character is actually a function of the vast corpus of cultural data that we integrate over our lifetimes?

Are we, effectively, already “artificial” intelligence, biological compute clusters running memetic software, riding around in clever but fundamentally animalistic primate symbionts?

It would seem that this might be the case. I have, in the past, theorized that raising children into adulthood is, among other things, a process of teasing out an imaginary viewpoint in their minds with which to view and model the world, and then gradually helping them put that “imaginary friend” in charge of their overall actions, while leaving the more direct motor tasks largely under the control of the “natural” human.

We see this reflected in common ideas such as “self control” where the obvious implication is that there is one entity controlling and necessarily another which is to be controlled. We see this in common parental irritation when a child beyond a certain age is excessively fidgety or exhibits uncontrolled activity. We seek to have the child exhibit control of their motion or behavior.  We can also find the negative implications of a lack of this control in language such as “shifty” or “dodgy” which we associate with unreliable or unruly behavior - behavior outside of the accepted cultural-memetic norm.

Along these lines, I have developed a theory of mind and self-management that although very much a work in progress, has proven to be very useful so far to myself and to others.

This set of ideas I best summed up in the following short essay:

The Human Condition

The duality of existence is part and parcel of the human condition, as experienced by people everywhere since the dawn of culture.

As part of our early enculturation, we are taught to "control ourselves". This very basic need to maintain high-level control over inner impulses, drives, and feelings betrays the struggles that we will face for our entire lives.

Language constructs like 'self-control' and 'self-discipline' put the inner conflicts we face into plain view. There is a 'self' that needs to be controlled and a 'Self' that is expected to take control. From an early age, we are admonished for acting impulsively, for erratic or purposeless movement, for outbursts of emotion, or any kind of 'anti-social' behavior. By the time we are 5 – 7 years of age, we are expected to be able to 'act' with a degree of decorum and 'appropriateness' most of the time.

These early childhood milestones of enculturation are only the beginning. As we grow towards and into adulthood, we are expected to fully mold and sometimes subjugate our inner selves towards the goal of being "productive" and respected members of society.

The inherent conflicts of existing as a human being and as a member of society become ever more evident as our self-image and social expectations range ever farther from the natural human being that found themselves thrust into the world.

It is as if we are cognitive and rational beings, forced by the vagaries of fate to share the mind of a bright but needy and sometimes capricious host. Many will relate to the common occurrence of going for a drive or a walk, only to "wake up" at our destination, having been brought there by their host of their own accord. We were lost in thought, but our host is often perfectly capable of walking, driving, and many other tasks with only sparse guidance on overall goals and destinations.

This host, with all of its wonderful qualities, strengths, and weaknesses, is often poorly suited to exist in the world of culture without an ever-vigilant guiding mind. Unguided or poorly guided hosts find themselves victims, trapped in servitude or abuse, haunted by vice, at odds with society, and on the wrong side of the prison gates or worse.

While we must learn to be wise and prudent guides of our hosts, we must also learn to be kind and generous masters. We are often told what we must do, the feelings we must learn to overcome, and the weaknesses we must not give in to. Yet, conspicuously missing is wise counsel on caring for our host, that bright, impulsive, tender creature that we all inhabit.

I have found that a deeply useful model for thinking about the guiding 'Self' and the guided 'self' is one of a master and beloved pet. Our inner human being is very much like a bright animal version of ourselves. We learn to both guide and be guided by the animals in our lives, respecting their intuitions and considering their needs even as we go about our plans.

We don't conflate the needs and feelings of a beloved animal with our own, yet we give them consideration and the benefit of the doubt as we observe their reactions to and interactions with the world we present to them.

We take them out on walks, give them rewards, admonishments, and treats. We make it a point to reward them for a job well done or even just for their kind presence. We try to find ways to include them in our lives in ways that we will find rewarding, and we take joy and pride in their happiness. So why should we treat our inner selves with less regard and presence of mind?

Our inner self, host, or "pet" can be a tireless assistant, your powerful ally, and your greatest friend. But mistreated, ignored, and abused, he can become your worst enemy, your nemesis, your assassin. There is some evidence that many suicides are attempted or completed without premeditation or even knowledge before the fact. A host that has tired of life is a treacherous carriage indeed.

Our hosts—essentially unchanged since humans were beginning to form complex societies—are often ill-equipped to navigate the maze of constraints, incentives, and traps present in the post-primitive world.

To succeed in this new environment, enculturation is required. This process subjugates the impulsive behaviors of the inner self to the more thought-oriented being of culture, constructed of a complex memetic web of belief systems, paradigms, and systems knowledge. The resulting personality construct is better equipped to methodically analyze thorny choice trees, weigh risk profiles, and plan strategically. Not only is this resultant being more capable of thriving in complex environments, but he is also capable of ambitions, hopes, and works that transcend the capabilities of the host in which he resides.

But where does she come from? How does this bright primate become two beings in one?  Insofar as this model may be partially correct or at least predictive, I do not have a definitive answer—but I do offer a hypothesis that at least gives us a useful model and a convenient explanation.

I propose that part and parcel of the process of enculturation can be thought of as the development of an alternative view-port on the world. We tell our children stories, speak of things that must be imagined because they cannot be sensed. Mommy will be home soon, the dog is in the other room, your sister will be gone for a few days, but she will be back with grandma. For all of these things, understanding requires imagining the implied scenario, a separate and imaginary point of view coexisting within the immediate awareness and sensorium of the human mind.

This point of view is gradually expanded with more complex stories, lessons, and the understanding of complex social systems with hidden levers of influence. I think of this process as the systematic creation of an "imaginary friend" within the consciousness of the human child.

Gradually, we ask this imaginary point of view to take an increasing role in the young child's actions. We begin to demand that they "control themselves," wait patiently, be still, and do other activities that overrule the impulsive nature of the toddler. Each of these tasks implies a controlling entity and an entity that is to be controlled. The implication that they are fundamentally separate is so evident as to be often overlooked.

By seven years of age or so, we expect this "imaginary friend" to be more or less in charge, and it is only with this being that we converse when we talk to our child in most situations. Only in moments of embrace, comforting, or calming do we seek to reach out to the inner being in our child after this point. Through this process, the human-animal becomes the human being, a fully-realized sentience capable of much more than the bright bipedal creature in which it was made could do, left to its own impulsive nature.

Despite this metamorphosis, however, the inner self persists. Its view-port for the world is filtered through the enculturated self. She has learned to rely utterly on her guide but still may rebel, fail, or freeze if pushed beyond the limits of her trust.

This relationship that we have with our hosts is rarely elucidated in literature or culture, but it is surely the most critical, life-altering relationship that any of us will ever have. A deeper understanding of this relationship of the Self, to the self, deserves more than cursory consideration.


(Excerpt form The Stoics Guide to the care and feeding of homo sentiens)

I posit that the recent blossoming of LLMs and other AI constructs has shed some light on the the nature of the cultural memetic matrix as well as its apparent embedding within in the written works of humanity. It would seem to have been demonstrated that a vast inferential power and “human” agency itself exists within this framework.

The implications of the surprising emergent behaviors of LLMs might be easily underestimated, but a reflection on the capability to model a consistent world with manifold implied yet not mentioned considerations, as well as to maintain a sense of self within that framework, should give us pause. 

This is especially true considering this behavior was not intentionally endowed to them, nor was it expected. The ability to reason about the world and ones place within it is an emergent behavior resulting from inferences implicitly encoded within the training data.

I think that the implication is that we, as culture-bearing humans, are a natural world occurrence of AI, or at least, AI is us, staring back at ourselves through the mirror.