I know this gets into philosophical matters. What is intelligence? What do we call intelligent? Can we even define, measure, or benchmark intelligence across entities with fundamentally different natures? For example, can we define and measure intelligence in dolphins using the same standards we use for humans? Or compare human intelligence with machine intelligence? I think there is a fundamental logical problem here. You cannot compare apples with hammers. They are fundamentally different things, and applying the same definition and measurement criteria to both doesn’t necessarily make any sense.
My definition of…
Let’s start with definitions, because that is where I have my fundamental issue:
Universal definitions tell us that intelligence is the ability to acquire knowledge and reason. To me, that is so abstract. Who even says that whatever the Oxford or Cambridge dictionaries say is accurate, relevant in all circumstances, let alone literally true? Obviously, those are just sources of knowledge, not something set in stone.
I mean, science changes, our understanding of things changes, and definitions should evolve with it. We need to revisit the definitions of words, put them into context, and consider what they actually mean in different circumstances, rather than simply relying on a dictionary definition as if it were an absolute truth.
But my problem with that definition is that knowledge itself is unclear, which makes the definition even more vague. What is knowledge, and what is not? If I read a manual and now know what is in it, is that knowledge? Seriously? So, basically, a script that can read through thousands of pages is incredibly knowledgeable? If I go to university and study philosophy, does that automatically make me knowledgeable about philosophy? What really is knowledge?
I say knowledge is more than simply knowing something. It is what you can figure out from what you know, particularly things that are not directly obvious from the original information.
Obviously, 2 × 2 = 4 is not what I would call knowledge, even if it is known by a three-year-old genius child. Knowing that, in Windows NTFS and ACL evaluation, an explicit Deny Access Control Entry always takes precedence over an explicit Allow ACE for the same principal and right, or knowing that the body relies on three primary energy pathways (the phosphagen system, the glycolytic system, and the oxidative or aerobic system), is not necessarily knowledge in the sense I am trying to describe. Knowing that software interrupt 21h is used by DOS to access operating system services doesn’t mean I know computers or understand how an operating system works. Memorizing every menu, feature, and service available in Microsoft Azure doesn’t make me knowledgeable about cloud computing. Neither does memorizing a list of the latest AI models make someone knowledgeable about artificial intelligence.
Those are things you know, regardless of when, where, or how you learn them. You could make the same three-year-old child memorize the NTFS access control mechanism and repeat it back to you. Does that mean the child is knowledgeable? Yet, most of the time, people will say someone is extremely knowledgeable about Azure simply because they know where every feature is located and what it does.
I may know something, many things, or even hypothetically everything, but that does not necessarily mean I understand what I know.
To me, knowledge is the ability to connect the dots, conceptually understand, generalize, or utilize what you know to explain other aspects of a system or science. Just knowing a protocol does not make me knowledgeable. Just knowing the entire book of law does not make me a lawyer, even if I can remember every single word of it. So, now it may be easier to define intelligence.
Intelligence from my perspective…
To me, intelligence is the ability to derive, discover, or create something previously unknown. Intelligence doesn’t necessarily have to start with previously acquired knowledge. It can also emerge from direct observation, experience, experimentation, and interaction with an unfamiliar environment.
Using the same examples above, intelligence is when, based on my knowledge of NTFS rules, I can independently troubleshoot an access issue that I have never encountered before. Intelligence is when I can correlate and explain the liver’s response to one of the metabolic pathways without having to read about that specific situation in a book or attend a medical lecture. Imagine someone dropped into the middle of the ocean without any prior knowledge of marine biology, oceanography, or even the environment itself. Through observation and interaction with that environment, they begin identifying patterns, understanding relationships, recognizing dangers, and figuring out how things work.
You won’t necessarily applying knowledge from books or previous studies. You are acquiring an understanding directly through your senses and your ability to make sense of what you experience. That, too, is intelligence.
Intelligence is the cognitive result of a transaction between multiple points of knowledge. It is when you connect those dots and shape something new out of them.
Intelligence creates something that did not previously exist. It may or may not use knowledge from the past but goes beyond it. It creates a new understanding, a new solution, or even a new skill. The birth of a neat idea is intelligence. This could be any piece of work, such as a book, a chapter of a book, or even a single sentence.
So, yeah, to me, originality goes hand in hand with what you want to define as intelligence.
If you create a work without external influence, that is one aspect of intelligence. But does intelligence really require creating something completely independently?
What is Not intelligence…
If I study Bach’s works for months and then, one Sunday morning, wake up and create a Baroque piece of music, am I really intelligent? I don’t think so. I have spent months studying someone else’s work, learning its structure, patterns, and characteristics, and then reproducing something based on what I have absorbed. Even if the resulting composition sounds original, that alone doesn’t demonstrate intelligence.
This is precisely where I see a parallel with what today’s machines do. They learn from enormous amounts of existing material and generate something based on the patterns they have absorbed. The result might look impressive, but the ability to reproduce or combine learned patterns should not automatically be confused with intelligence.
Intelligence can create skills that did not previously exist, or it can combine a set of existing skills and create a new one. If I learn assembly language and then learn Python much faster and more easily because of what I already understand, is that an indication of intelligence?
Now, let’s bring AI into the picture.
We give a machine five years of stock market data and tell it to follow an algorithm to predict what “we think” is hidden in that data. Is that machine intelligence, or is it the creator’s fascinating concept of deep learning?
Did AI create itself? Was there ever a piece of code running one day that independently created Transformers? Those who developed Transformers, those who utilized them to build LLMs, and those who invented the underlying concepts demonstrated intelligence. But does that automatically mean the resulting machine is intelligent? This is where I think we need to make a distinction.
In the context of my article, artificial intelligence is the knowledge, creativity, and engineering behind the creation of AI itself, its subsets, and concepts such as teaching machines to learn (aka machine learning). These inventions are manifestations of human intelligence. The machines built using these inventions may eventually demonstrate artificial intelligence, but whether they already do is a completely different question.
I am simply trying to say that what the mainstream defines as artificial intelligence is not necessarily accurate because we don’t even have a sufficiently clear understanding of knowledge, as I explained above, let alone a universally accepted definition of intelligence in the first place.
What these machines are doing…
That so-called intelligent model we interact with is, in my view, a machine-based manifestation of an enormous amount of human knowledge. It has absorbed patterns, information, relationships, and representations derived from what humans have produced. It certainly doesn’t contain everything humans know, but it contains an incredible amount.
I wish they had called it artificial whatever, or generally a smart machine, but not necessarily intelligent.
What we use today, which, by the way, is fascinating, no doubt, is something much more sophisticated than a traditional knowledge base. It is a system capable of processing enormous amounts of information, identifying patterns, connecting concepts, and generating things that may not have existed in exactly the same form before.
I mean, imagine that back in the day, you probably had to spend five years of your life studying Bach’s works before you could truly mimic him. But with today’s hardware and software, which are very cheap and accessible, even in your own home, you can train a machine to compose music resembling Bach’s style, probably in less than two weeks?
You may controversially argue that AI is capable of generating a piece of Baroque music without even coming close to Bach. A machine can generate music by combining patterns and elements from multiple, sometimes completely unrelated, sources. But that is just pattern mixing, almost random, and perhaps pure luck. It is not the same as the actual evolution of music from Renaissance to Baroque, which involved a gradual development of musical ideas, techniques, and understanding. And that distinction matters. That is not how computer science itself evolved to the point where we could practically apply mathematics to teach machines how to learn.
Intelligence is when you land a spaceship on the Mars with zero error tolerance. Intelligence is Einstein’s theory of relativity. Intelligence is the emergence of social media: the idea, the utility, the mindset, and the implementation behind it, not simply the machines that make it work. Intelligence is the discovery of a new element. Intelligence is understanding the universe.
What we have today is a breakthrough in the way humans make mathematics a reality of everyday life. It is fascinating, and there is an enormous amount of intelligence behind it. But that does not automatically mean a chatbot writing a super-cool story for you or automating your online shopping is intelligent.
I call these smart machines. Computerized information processing systems, or computerized experience machines…well, call them whatever you want. Anything except intelligent!