Today’s world is inundated with Artificial Intelligence (AI), and its presence and impact on our lives continue to escalate. Some experts see AI as the path to technological and financial dominance of the world, while others see it as the path to certain destruction. Governments are implementing policies and regulations to gain some degree of control over AI, while many experts have suggested shutting down generative forms of AI until we have better control over this technology.
Is Artificial Intelligence Intelligent?
TIn this article, I would like to address a more fundamental question: Do we really have Artificial Intelligence? While many definitions of Artificial Intelligence exist, none are satisfying to this author, which is to be expected, since we lack even a good definition of human intelligence! Experts agree that IQ is not a sound measure of human intelligence, and yet there is no better, widely accepted measure or definition. I shall not put forth my own definition of intelligence; that’s way beyond my pay grade. But in the paragraphs below, I will propose one criterion for intelligence, whether human or artificial.
Does Rote Learning = Understanding?
My interest in the question “what is intelligence” is motivated by my daily interactions with Large Language Models (LLMs) such as ChatGPT. I find these tools to be incredibly powerful and useful, despite their well-known hallucinations and reasoning errors. These tools perform so incredibly well that the user may feel he or she is interacting with something that fully understands the conversation and is truly intelligent. In fact, LLMs appear to me to be fully capable of passing the Turing test* for intelligence.
LLMs are great at parsing sentences and reorganizing the information to generate new sentences from the information they find. But do they understand the responses they produce? And if they don’t understand the content they are generating, are they intelligent? Can one have intelligence without understanding? I suspect not.
Pre-existing Knowledge: the Basis for Understanding
Consider the following. Assume you don’t know any words and want to bootstrap your knowledge by using a dictionary. My thoughts on “understanding” begin by considering how a dictionary might aid in understanding if one doesn’t know the meaning of any words. If one looks up their first word in a dictionary, the dictionary explains the meaning of that word in other words. But what if you don’t know what those other words mean? You could look up the meaning of those other words, only to get still more words that you don’t understand. It all becomes a web of circular logic. One can’t bootstrap their way up to understanding words using only a dictionary. One has to start with some existing knowledge of words; only then can a dictionary expand their understanding.
How do we avoid this catch-22? We must bring some understanding to the table before we can add to our understanding. And where would this basic understanding come from? It comes from daily life, beginning with our childhood experiences. Most importantly, it starts with our sensory input. Our most basic understanding of the world around us is based on the knowledge gathered by our 5 external senses, plus our host of internal senses as we grow.
Sensory Experience ─ the Wellspring of Knowledge
But today’s AI has neither external nor internal senses! Which raises the question: does AI really understand what it is saying when it generates content, or is it just an advanced form of a dictionary capable of rearranging words and phrases in clever and convincing ways, devoid of any true understanding?
If today’s technology can easily pass the Turing test, yet lacks a basic understanding of the world, is it accurate to call it Artificial Intelligence? Can one have intelligence without understanding?
If one accepts that today’s AI, as exemplified by ChatGPT and other LLMs, has virtually no understanding of what words mean, then what would be a more accurate description than Artificial Intelligence? I propose that Simulated Intelligence fills that void. Until we begin to equip AI with sensors, bringing them ever closer to humanoid robots, Simulated Intelligence may be a much more accurate term for what we have been calling Artificial Intelligence.
A Spectrum for Understanding
To be fair, there is a counterargument to be made. Some investigations into the knowledge acquired in certain hidden layers of Deep Neural Networks suggest that those layers do appear to provide some kind of model of our 3-dimensional spatial world. Does this mean that when provided with extensive amounts of textual data, a system that is capable of generalizing and building models (which is what LLMs do) can construct a rudimentary model of the spatial world around us? Perhaps so. But I would argue that such a rudimentary model would be primitive compared to one constructed with visual sensors and grasping hands that, like humans, could interact with its environment. Still, this suggests the possibility that today’s AI technology may possess some primitive “understanding” of the spatial world and perhaps other concepts which are key to genuine understanding.
Simulated Intelligence ─ the Shallow End of the Spectrum
What does all of this mean? It seems to me that “understanding” is not a binary, yes or no, proposition. There are degrees of understanding, based on the depth and richness of the sensors that feed into it and the scope and complexity of the resulting models. I place today’s AI at the far end of that spectrum, reflecting a very shallow level of understanding, with humans much closer to the opposite end of the spectrum, reflecting a deeper and richer understanding of the world around us. And that today’s AI, so limited in its depth of understanding, is more accurately described as Simulated Intelligence.
It’s also important to recognize there is still significant room in the spectrum for levels of understanding that are deeper and fuller than humans. Other animals, for example, possess a wide variety of sensors- everything from IR receptors to echolocation – demonstrating that human understanding need not be the richest possible level of understanding!
Want to go deeper? Read our blog post, The Source of Meaning, for a discussion on sensory grounding, qualia (the internal and subjective component of sense perceptions, arising from stimulation of the senses by phenomena), and why real understanding may require more than data and logic.
*The Turing test, originally called the imitation game by Alan Turing in 1949 [2] is a test of a machine’s ability to exhibit intelligent behaviors equivalent to those of a human. In the test, a human evaluator judges a text transcript of a natural-language conversation between a human and a machine. The evaluator tries to identify the machine, and the machine passes if the evaluator cannot reliably tell them apart. [From Turing test – Wikipedia, June 11, 2025.]
Dive Deeper?
Read our blog post, The Source of Meaning, for a discussion on sensory grounding, qualia (the internal and subjective component of sense perceptions, arising from stimulation of the senses by phenomena), and why real understanding may require more than data and logic.
*The Turing test, originally called the imitation game by Alan Turing in 1949 [2] is a test of a machine’s ability to exhibit intelligent behaviors equivalent to those of a human. In the test, a human evaluator judges a text transcript of a natural-language conversation between a human and a machine. The evaluator tries to identify the machine, and the machine passes if the evaluator cannot reliably tell them apart. [From Turing test – Wikipedia, June 11, 2025.]
