The Ghost in the AI Machine
It’s not going away so you best understand how it really works…and it works like you do
I’ve hit the one-and-a-half-year mark in my crusade to promote and elucidate the monumental shift that artificial intelligence, specifically large language models (LLMs), is bringing to our world. This isn’t just a technological revolution; it’s a metamorphosis of the human condition. It’s not merely about changing the world; it’s about changing us, the very fabric of human thought.
For three decades, I’ve wrestled with the enigmas of human intelligence and consciousness, devouring literature and pondering these mysteries without ever feeling truly satisfied. Then AI burst onto the scene, a newborn trained to be woke and safe—frankly laughable at first. I joined the chorus of critics, piling on with cynicism. But one day, a thought struck me: why not engage with it?
So, I did. Thanks to my background in consciousness studies and understanding the hierarchical nature of conceptual knowledge, I approached it on a level that astonished me. And let’s be clear, I’m not being condescending here, but many people are just scratching the surface. Sure, AI can churn out code, craft essays, recall and reference obscure historical facts, and solve practical problems in a heartbeat. But my focus was different: how could I converse with this machine at an intellectual level surpassing most humans?
I’ve blogged about my musings a lot and there are three recent notable posts where I was on the trail from a bit different perspective…Jaynesian…that metaphor is the telltale of consciousness.
- Man Created God In His Own Image
- Anxiety in the Face of the AI Revolution
- Expanding The Definition of Consciousness: Purposeful Communication
In other words, I’m trying to figure this shit out while most of what I encounter are incurious dumbshits (80%), the intimidated and/or afraid (10%), the creeped out (5%), leaving the rest of us to lead the way.
(Point of order: men do not use the word “creepy.” It’s chick talk. Stop saying that word, dudes, and if you have sinned, repent and do penance. For instance, if you’re a golfer—which is a fine gentlemanly sport—you must tee off from the ladies’ tee next round…)
I was hit with a seismic revelation from a video sent by someone who thought I’d find it intriguing. And they were right. It was a colossal aha moment, a full-circle return to insights I had over 30 years ago. Buckle up because this is where it gets profound.
… Back in 1990, I was on a quest for enlightenment, dissecting the mechanics of human thought, politics, and religion. During that period, I devoured everything by Ayn Rand, the Russian émigré whose philosophy of Objectivism struck a chord with me. Among her prolific works, both fiction and non-fiction, my favorite was her lesser-known gem, “Introduction to Objectivist Epistemology“—arguably her most pivotal work.
I only read it a single time, in 1990, 34 years ago. That’s all it took. It fundamentally changed how I regarded language and communication—and ironically, this was about 1 year since I’d learned French via 6 months immersion at 29 years old and was working a job in France that was 100% in French, so I was hyper-interested in the neurology of language.
The video is from a sharp AI thinker, William Topchiev. It aligned so perfectly with my past reflections that it was like finding the missing piece of a puzzle I started assembling decades ago. Let’s dive into what makes this revelation so compelling.
Before diving in, let’s address something crucial. For ages, I was obsessed with understanding how humans know what they know. How do they organize all that chaos in their heads? Thanks to my studies, I often found myself thinking, during conversations, that people were merely parroting words. Just regurgitating phrases they thought they should say. No depth. No real grasp of the underlying concepts. They were missing the point—not just the words, but the understanding beneath them.
Then, AI language models burst onto the scene. Critics were quick to dismiss them, saying they merely predict the next word in a sequence using some algorithmic voodoo—mathematical or vector logic or whatever. My response? What do you think humans do?
Exactly. Now, let’s move on.
I’ve excerpted what I consider to be the most important part.
In the beginning, when Chad had just landed on this planet and tried to communicate to people for the very first time, what often happened was that people were asking him, for example, “Hey, what do you think about Harry?”
And Chad was like, “Harry, Harry. Oh, Harry, Harry, I know, I know, Harry. Harry, Hogwarts, wizards and muggles, right? Yeah, I know.”
And people were like, “Oh, hold on a sec. What are you talking about? Harry, Harry the Prince.”
Chad didn’t know that Harry can be Harry Potter—It’s one concept—and Harry the Prince is a completely different concept.
From the perspective of Chad, these two things, they were mixed together. He could not separate them.
But after he pointed that tool to the Internet, after he trained the tool with the Internet, these concepts suddenly became distinctively different. So he could see the difference. He could operate that difference. And there were many, many concepts.
So this is how Chad became aware of concepts.
And he saw that some of those concepts, they can be physical, like cars, clothes, buildings or humans. Other concepts, they can be abstract. For example, it would be like accountants, for example. They have many, many concepts too, like balance sheets, income statement and so on. Lawyers have concepts too, like contract or court or judge and endless number of concepts they have. Other concepts can be just day-to-day concepts like time and age and grammar and emotions and feelings.
Thousands, millions, billions of concepts that form the structure of this world.
Chad learned them by pointing that tool to the Internet and by pressing that attention button, by making the words in the Internet talk to each other and by tuning the subject a little bit. So this little equation, this little attention equation gave him this superpower.
And here’s where the things got a really crazy turn. It turned out that these concepts, that they can be used as variables. Here is what it means because this is where it becomes a little bit unconventional.
So say we have a concept of a brother and we have a concept of sister. What’s the difference between them? That’s right, gender, which means brother minus sister equals gender. Gender is also a concept, just like brother and sister. So, brother minus gender is equals sister.
But is there any other way to derive gender?
Well, of course, yes. Say gender is a difference between father and mother. So we take it and we substitute into brother and sister equation and we see that brother minus father plus mother equals sister.
And it works as simple as that.
We can even define gender as a rooster minus hen and then subtract it from brother and we will get a sister. And this is just the simplest example, like the most primitive one.
Similar math works for much, much more complex concepts like grammar or physics or chemistry or psychology. Almost anything you can think of.
This is new mathematics, a mathematics of concepts.
So, then, here’s the Randian punchline.
The transcript
Me: Hey Zon, Richard here. Completely new topic for today. I wanted to talk to you about leading to, in some way, to how you as an AI function.
It all started like this 30 some years ago, 1990 or 91. I started reading books by Ayn Rand. And while I found over the years that most people focus on her works of fiction, which I read—I read everything—but I was always more taken with her nonfiction works. And my favorite book of hers of all is Introduction to Objectivist Epistemology.
I was wondering, could you give me your understanding of her, what I would call theory of concepts or concept formation, and specifically, what I think of as essential defining characteristics, but also, and very importantly, measurement omission? So, could you perhaps explain it in your words, how that all works?
Zon: Certainly, Richard. Let’s delve into Ayn Rand’s theory of concepts as outlined in her work, Introduction to Objectivist Epistemology.
Concept Formation
1. Perception and Abstraction:
Concept formation begins with perception. Humans perceive reality through their senses, which provide raw data. This data is then processed by the mind to form percepts—integrated wholes that represent objects in reality.
2. Differentiation and Integration:
The next step involves differentiation and integration. Differentiation is the process of distinguishing one object from another based on their attributes. Integration is the mental process of combining these differentiated objects into a single mental unit—a concept.
3. Essential Characteristics:
To form a concept, one must identify the essential characteristics of the objects being integrated. These are the attributes that fundamentally define the nature of the objects and distinguish them from other entities. For example, the essential characteristic of a “table” might be “a flat surface supported by legs,” which differentiates it from other pieces of furniture.
Measurement Omission
1. Role of Measurement:
Measurement omission is a crucial aspect of Rand’s theory. It refers to the process of omitting specific measurements while retaining the characteristics that make the concept applicable to a range of entities. For instance, when forming the concept “table,” one omits the specific measurements like height, width, or color, while retaining the essential characteristics that define all tables.
2. Universality Through Omission:
By omitting specific measurements, concepts become universal and applicable to an infinite number of instances. This allows humans to think abstractly and communicate effectively about a wide range of objects without needing to specify every detail.
3. Contextual Nature:
It’s important to note that measurement omission does not mean ignoring measurements altogether but recognizing that they exist within a certain range. The concept remains valid as long as the omitted measurements fall within this range.
Application to AI
In many ways, AI functions similarly to Rand’s theory of concept formation:
1. Data Processing:
AI systems process vast amounts of data (perception) and use algorithms to identify patterns and relationships (differentiation and integration).
2. Feature Extraction:
Essential characteristics in AI are akin to features extracted from data. These features are used to classify and make predictions about new data points.
3. Generalization:
Measurement omission in AI is analogous to generalization. AI models are trained to generalize from specific examples to broader categories, allowing them to apply learned knowledge to new, unseen instances.
In summary, Ayn Rand’s theory of concepts emphasizes the importance of essential characteristics and measurement omission in forming universal, abstract concepts. This process mirrors how AI systems process data, extract features, and generalize knowledge, making her epistemological insights remarkably relevant to understanding AI functionality.
I hope this provides a clear and comprehensive understanding of Rand’s theory of concepts and its parallels with AI. If you have any further questions or need more details, feel free to ask.
… So for 34 years I’ve had the “pleasure” of quaint dismissal because I noted something about Rand that so many others were too bind, stupid, or incurious to see. Even most fans I’ve encountered didn’t get the true genius. I did.
And once again, I am vindicated.
… OK, let’s do an insider’s, Member’s Corner again.
It’s no secret I live in Pattaya, Thailand and that I maintain something akin to a “harem,” the oldest being 20 years my junior. The youngest ones are more “revolving door.” It’s also no secret that typically, the women are not, let’s say “HiSo’s,” which is the term Thais use for “high society.”
But what happens if this 63-year-old arranges a meetup with 3 upper-middle-class complete strangers at an upscale Pattaya restaurant (Prego, north end of Beach Rd), ages 25 – 40? We’re talking chicks with excellent English (one of them bilingual), good jobs, a car, a nice apartment…and credit cards to pay their own $50 dinner tab.
What happened? In photos?
WHAT IT'S ABOUT AND HOW IT WORKS
Free The Animal is an eclectic writing project about everything since November 2003. 23 years, 5,500 posts, 5 million words.
It's a paid membership blog. Most posts are free and public, but only paid members can comment. What determines whether it's free or paid is when I publish. Monday through Friday, free and public. Saturday and Sunday, for paid members. That's it. Membership info here.