The latest BIG idea in western culture is so-called ‘Artificial Intelligence’. Some view it as the answer to all of our problems, others a threat to human survival. Its neither of these. But it does have serious implications for the discernment of truth and our political future. The main concern with it is the fact that Large Language Models (LLMs) have achieved the ability to fool people into regarding them as though they are human. People falsely attribute to them sentience and agency, some even seeing them as ‘living beings’. To understand the real threat that so-called “AI” poses, we must answer the question, how much of the AI narrative is realistic, and how much is just hype?
[In this context, sentience, means the ability to perceive or to feel things, as opposed to being fed, or accessing, data. Agency, means taking self initiated, as opposed to automated, actions toward a goal on behalf of someone else.]
What is AI?
The term “Artificial Intelligence” is a misnomer right from the start. It’s a marketing term that doesn’t refer to a single thing, but a collection of computer based software tools that can perform a variety of automated tasks, from spell checking, to conversational style interactions with Large Language Models like Chat GPT. But it’s a very clever marketing term that immediately implies that machines can be intelligent. This sounds impressive, but there are no facts to support it. And of course, it all rests on how we define intelligence. But the marketing term has worked, because the idea of computers being intelligent, and therefore, potentially smarter than humans, is now firmly entrenched in most people world view.
According to Emily Bender and Alex Hanna, in their book “The AI Con”, So-called “AI” can more accurately be thought of as various computerised automation systems that deal with several different types of automation. These include automatic decision making systems that are used for things like approving loans, or allocating social benefits. Then there is automatic classification, which is used for things like sorting and matching images, for surveillance or personal photo use. A third type is automatic recommendation, which we see on our Youtube or social media ‘feed’. The fourth type is automatic translation from one format to another, such as transcription (speech recognition) or text to speech, as well as automatic language translation. And lastly, there is automatic text and image generation. This is the kind that is being talked about a lot at the moment. This kind of automation allows people to generate images, video, or plausible sounding text, based on textual prompts. All of these software capabilities are a part of what is collectively referred to as ‘AI’. And lumping them all together supports the illusion of ‘intelligent’ technology.
Definition of intelligence?
Definitions are very important. They tell us what a word means. And generally speaking when we understand what something is, we hold an implicit definition in our mind, even if we have never put it into words. And when we genuinely know what something is, we know what we are talking about when discussing what it can and cannot do.
In general, a definition is a reflection of the common understanding of a conceptual idea, but that understanding should properly be grounded in facts—and generally speaking, it is. But if the way a word is used changes, its meaning can change. And if the use of a specific word, or phrase, is changed deliberately by, for example, a marketing term, or by propaganda, then the meaning/definition can be effected, even changed by the new usage. This is what is happening to the common understanding of intelligence as a consequence of people discussing “artificial intelligence”.
I used GROK as a search engine to find definitions of the word ‘intelligence’. The output was as follows.
“Intelligence lacks a single universally agreed-upon definition, as it varies across philosophy, psychology, biology, and other fields. It generally involves capacities like reasoning, learning, problem-solving, adaptation, and abstract thinking, but emphases differ by era, culture, and perspective.”
I use the definition of intelligence that I remember from my education some 45 years ago.
‘Intelligence is the ability to think with abstract conceptual ideas’
Although not universally accepted, this is a fundamental definition because it is this ability to use (or reason with) conceptual ideas that makes the application of knowledge to problem solving possible. And it’s the application to problem solving (in all sorts of contexts) that is the common denominator of the definitions I found using GROK.
If we focus on the fundamentals, we can see that intelligence can only be an attribute of a conceptual form of consciousness. More specifically, only of a living organism that possesses a conceptual consciousness—which means a human being. Whatever animal lovers may like to think about their pets, they cannot be ‘intelligent’ according to this definition. But this doesn’t mean that they cannot learn things and use knowledge, because intelligence is not about these things. Intelligence is a measure of the human ability to use (or think with) conceptual ideas—both in number, and in degree of abstraction.
Machines and Humans
When a machine is capable of solving a problem or performing a task, it is a consequence of the application of the intelligence of the human conceptual consciousness that designed the machine to perform that task. Intelligence is not an attribute of a machine that flawlessly performs particular tasks, even if the word ‘learning’ is used to describe its data accumulation and storage. This is because LLMs (or any other automation software applications) have no understanding of any concepts. We must remember that all words (except proper nouns) are concepts. The fact that computer algorithms are excellent at pattern recognition, and can do many tasks much better than humans, including modelling how we use language, is not an argument that they are intelligent.
According to Professor Michael John Wooldridge, the learning that an LLM does is not how humans learn language, and its not how language evolved. This is because there is no conceptualisation going on. The debate today ignores how human consciousness forms a concept with understanding behind it, understanding that relates the word representing the concept back to the real world. Knowing how words can be associated with other words is only part of the picture of ‘using’ language. The all-important foundation is holding an understanding of a concept that relates back to experiential knowledge in reality.
Most of the time when people refer to “AI”, they mean Large Language Models, or LLMs, like Chat GPT, Claude, Perplexity, or GROK, etc. So what exactly are these LLMs?
Large Language Models (LLMs)
The technology of language models began with the simple n-gram models developed in the 1940s. In these models “n” stands for a number, and “gram” is a word. When fed with a body of text, the basic unigram model (where n is one) can count up and rank the frequency of each word. The next simplest is a bigram model (where n is two). The calculations now consider pairs of adjacent words, and predict the likelihood of any two particular words appearing together. For trigram models (where n is three) the calculations are made over strings of three words. For each pair of words, the next most likely word is calculated from the training data. This n-gram model can be extended to larger values of n. The predictive text on your phone uses similar technology.
Then came the so-called “neural” language models. These use algorithms called “neural nets” which are composites of mathematical functions called “perceptrons”. But, once again, we must remember that these are carefully chosen terms that sound impressively similar to how the human consciousness might work. If a computer uses a “neural network”, surely it’s thinking, right? Wrong! An LLM using neural networks, is essentially a ‘next most likely word’ predicting device. As the amount of ‘training data’ increases, they get better and better at predicting the next most likely word. But this is most emphatically NOT the same as thinking, even if it looks like thinking. In their book “The AI Con” Emily Bender and Alex Hanna say “Think of the output display of ChatGPT as an illusion, or a magician’s tool.
Language models model language. From the source training data, they ‘learn’ how humans use words and then they mimic a human using words. As they get better at knowing which words and phrases most likely go with which others, they get good at appearing to be like a human being. Why? Because this is what human beings do, and only human beings do it. Remember that we humans are the only living organisms on Earth that possesses a conceptual form of consciousness—which means we are the only living organisms that use words/concepts to categorise what we experience in the real world into knowledge that can be used for problem solving. Language is the key distinguishing feature of humans, but it is not yet known precisely how we form the concepts that make up our language, or how we imbue them with meaning. LLMs model the way we use language, but that is all they do. They do not think, they do not reason, they do not understand. They do not possess consciousness. They are not sentient, and are not capable of taking independent action on their own initiative. We must remember that if it is not yet known how we humans form our concepts and imbue them with meaning, then it is not credible to suggest that software programmers have figured out a way to do it in a computer. And to be fair, no one is claiming that they have. But if they had, this would take us much closer to a genuine computer intelligence than what is being discussed today.
Conceptual Consciousness
When we use language, we choose a word that matches the meaning we wish to convey. Words are the building blocks which construct the meaning we wish to share with another conceptual consciousness. Each specific word has a specific meaning ascribed to it. For example, “The—cat—sat—on—the—mat.” Broadly speaking, we understand the meaning of each component word (concept) and assemble words to construct any potentially complex message we wish to communicate. But LLMs have no understanding of meaning. They do not form concepts. This is why it is nonsense to suggest that they are “very like us”, as Geoff Hinton does. He is wrong, even if he is a Nobel Prize winner.
The upshot of all this is that due to the way our human form of consciousness works, we tend to over estimate, and attribute agency and intelligence to the external world and to other entities. Most obviously with dogs and dolphins, but also now with computers. We imagine intelligence where there is none. We readily attribute human-like qualities (those of a conceptual consciousness) to ‘the source’ of conceptual ideas which are the output of LLMs like Chat GPT. We know that when we hear a string of conceptual ideas in a recorded message that the machine playing the recording isn’t the source of those ideas, it was the mind behind the voice. But when we play with chat bots we are easily fooled into thinking their output is a string of conceptual ideas from a mind, like us. But the chat bots and LLMs are modelling our use of language. They are computers pretending to be human, and the latest software systems are very good at it.
Anthropomorphising Computer Software
To ‘anthropomorphise’ means to attribute human characteristics or behaviour to something—a god, an animal, or an object. Almost all of the language used to describe “AI” and its functionality encourages us to anthropomorphise. At the end of this post you will find a list of such terms, as well as some more appropriate suggestions to use instead.
Who benefits from “AI” Hype?
It isn’t common to ask the question Cui Bono? (who benefits?) But it should be one of the first questions asked by the truth seeker when considering any ‘high profile’ and/or ‘high consequence’ ideas in the mainstream narrative. Who benefits if everyone thinks AI can do much more than it really can? Who would it serve if everyone believed AI was sentient, or conscious? The answer is the same when we apply this question to any false idea. Who benefits from someone else believing a false idea to be true? The answer is, those who wish to deceive. This is a very important point, and it highlights why we must be very careful to understand precisely what “AI” is, and what it is not.
Clearly there are those who have a vested interest in people believing an exaggerated version of AI’s potential. The obvious example is that funding from venture capitalists is more likely to flow to ‘start ups’’ if they inflate and exaggerate the estimates of future functional capabilities. The recent Initial Public Offering (IPO) of Anthropic, and Open AI, saw valuations riding high on the all the hyped expectations and promises of AI. What we are seeing here, is money flowing into research and development of the industry that is being created to monitor, track, and surveil us, on the basis of exaggeration and hype.
Political Usefulness
But beyond the financial aspect of this, a false understanding of what “AI” is, and of what it can do, can be politically useful. And as truth seekers, we should remember that political motivations are always more fundamental than financial ones. If we zoom out for a moment. The world is controlled by the owners of the central banks (whoever they are). Money is important, but it is already plentiful for them, and is not their primary motivation. Their deeper motive is the establishment of a global collectivist political system. And this agenda would benefit enormously from the development of computerised automation systems. The Hype helps money flow into the research and development, including yours and my money in the form of pension funds and other investments. The hype pushes and powers the research and development. And this is in the interests of those who seek to establish a global collectivist political system.
But the point is that it’s not all about money. Consider the following. If you think that a chat bot is intelligent, and is possibly conscious, you may inappropriately place greater trust in it than you should. This may cause some people to seek the advice of “AI” to make a decision on an important issue, on the assumption that ‘it knows more than we do’. The belief that AI is intelligent, and that it actually thinks, opens the door for chatbots to encourage people to vote one way or another, or to believe in some specific ideas, and reject others. The bottom line is, that unless you are fully conscious of the fact that you are communicating with a piece of software, NOT a living being, you are likely to be deceived. The trap is to fall for the idea that the machine (the software) is a thinking reasoning entity. Its not! But its very good at appearing to be.
We should also consider who is programming these LLMs, and for who’s benefit. We must not make the mistake of regarding them as impartial! They can easily be set up to be biased towards the promulgation of particular ideas, and they very likely are.
Another concern is that people who think AI is conscious may start to want to grant rights to it! This has potentially serious consequences for our already confused understanding of what rights are. The concept can only properly pertain to a volitional conceptual consciousness, not to machines, or even lower animals. The concept of ‘rights’ is our means to freedom. And, as such, it must be carefully guarded. (see here for more on rights).
In the 1939 film “The Wizard of Oz” there is a scene where Dorothy and her companions (the tin man, the lion, and the scarecrow) confront the Wizard. During their conversation with the Wizard, Dorothy’s dog pulls back a curtain, revealing a little man who is speaking into a microphone and pulling the levers that operate the machine that is the Wizard. As his booming voice and all the thunderous effects give the appearance of a Wizard, the reality they discover is that it is all a charade! The truth was, that Dorothy and her friends didn’t need the magic of the Wizard, they already possessed the means to solve their own problems.
Think about this. And consider the following imagined thought process of those misleading us towards global collectivist governance.
“If we can instil in the minds of people a misplaced ‘trust’ in the computer systems that we control by referring to them collectively as ‘Artificial Intelligence’, then we will be able to fool people into accepting ideas that they otherwise would not accept if they knew that those ideas came directly from us!”
The Negative Consequences for Truth and Freedom
There are serious implications of so-called “AI” for our ability to discern the truth of what is going on in our world.
LLMs are essentially ‘most likely next word’ predictors. We know that they get better and better at this task with an ever increasing amount of training data. So there is an incentive to expand the training data to include everything on the internet, if this hasn’t been done already. But the consequence of this, is that the explanation of any idea, event, or phenomenon, offered by an LLM will simply be the most frequently published explanation on the net. LLMs will go with the statistically most likely. In terms of truthful information this ‘method’ results in the computer presenting what will be interpreted as truth on the basis of majority view. The upshot of this is that people may believe the ‘clever’ computer is giving them truth, when in fact, it’s giving them either the majority view, or, the ‘officially sanctioned’ answer to their question pre-programmed into the LLM. Asking “AI” will either offer ‘truth’ by numbers, or just add another layer of censorship and deception.
We cannot trust computers to tell us the truth, no matter how intelligent or clever they are presented as being. A computer, or the software that runs on it, is a machine, and like the Wizard in the film Wizard of Oz, it has specific human creators and operators who direct it, and who have their own agenda. There is a danger that “AI” becomes a proxy for the people behind the ideas that “AI” recommends, tries to sell us.
And then we must consider that pattern recognition and automated classification systems are very useful in spotting anomalies, or unusual behaviour, across a large data set. This means it is very good at identifying dissenters and resistance to the agenda. It means that anyone doing things differently, can be identified and placed under closer scrutiny if/as required. Companies like Palantir have been pioneering such software, and are already using it. This is perhaps one of the most disturbing uses of so-called “AI”, due to its implications in assisting tyranny.
The Turing Test
In his 1950 paper “Computing Machinery and Intelligence”, Alan Turing proposed a thought experiment on the subject of machine intelligence. He avoided the awkward question of ‘can machines think?’, and instead posed a different question. He asked, “Can a machine converse so convincingly that a human cannot reliably distinguish it from another human?” with the implication that if it can, then the machine must be intelligent. If you think about it, this is a cunning intellectual manoeuvre to shift the debate away from ‘what is thinking?’. Instead he proposes that we should conclude that the computer can think, and is indeed intelligent, if it can fool us into thinking it is human. Effectively, the Turing test has been passed. Because people are fooled. But, in fact, passing the ‘test’ means nothing. The test is so meaningless it should never have been given the credibility that it has. Why do so many commentators give credence to this test by even mentioning it?
Interestingly, Alan left an extensive legacy in mathematics and computing, but as a homosexual, he had no descendants, only one brother. He died by suicide aged just 41. I wonder if he had served his usefulness and had gone ‘off-script’? Or perhaps his legacy was desirable to some criminal parties?
Conclusion
Understanding precisely what AI is and what it is NOT, is the biggest truth issue of today. We are being encouraged to accept a false understanding of what is disingenuously described as “Artificial Intelligence”. This false understanding has serious negative consequences, just as the belief in any false idea has negative consequences. It causes money and effort to flow into developing the tools needed for an oppressive Technocracy, effectively getting us to pay for the construction of our own prison. It also presents an opportunity for those who would mislead us, into attributing ideas they wish us to accept, to a supposedly intelligent agent, so that those ideas will be more likely accepted. And it provides a seemingly impartial tool for people to use in researching the truth, that is in fact simply a presentation of the majority (mainstream) view, or the desired propaganda message.
I am not saying that AI is not useful. But I do think we should exercise extreme caution in using it, because ultimately, it is the tool of those who wish to enslave us. As we interact with “AI” it gets better and better at telling us what we want to hear. It learns about us, from us.
Properly understood, the technology that comprises “AI” is something that augments rather than replaces our intelligence. But in the current socio-political context, with such widespread misunderstanding of its nature and its capabilities, so-called “AI” is a threat to truth and freedom.
What do you think about “AI”? Join the conversation and leave a comment
Some interesting videos
DAWKINS himself on ‘Could Claude be conscious?’ – https://archive.is/XWdlT
Gary Marcus on Richard Dawkins and the CLAUDE DELUSION – https://garymarcus.substack.com/p/richard-dawkins-and-the-claude-delusio
Gary Marcus on the massive problems facing AI – https://www.youtube.com/watch?v=aI7XknJJC5Q
Anil Seth on why AI will never be conscious – https://www.ted.com/talks/anil_seth_why_ai_isn_t_going_to_become_conscious
Anil Seth on “Within Reason ” with Alex O’Connor, discussing why AI will never be conscious – https://www.youtube.com/watch?v=lsi8T_WtLnE
Strong anthropomorphic terms
- Hallucination: describes false or unsupported output as if the system were having a perceptual disturbance.
- Understanding: implies comprehension, when the system may only be producing contextually plausible language.
- Reasoning: implies deliberate thought, when it may be pattern-based generation or formal step production.
- Thinking: implies inner mental activity.
- Knowing: implies possession of knowledge rather than access to encoded statistical patterns or retrieved data.
- Learning: implies education or experience, especially misleading when used about a single chat.
- Memory: implies personal recollection rather than stored context, logs, vectors, or user preferences.
- Forgetting: implies a mind losing memories rather than context being truncated, unavailable, or not retrieved.
- Attention: sounds like conscious focus, though technically it refers to weighted relationships between tokens.
- Creativity: implies imagination or inspiration, rather than recombination and generation within learned patterns.
- Imagination: suggests mental imagery or fantasy.
- Intelligence: implies mind-like capacity, though the word is doing a lot of philosophical work.
- General intelligence: suggests a unified, human-like intellect.
- Emergent behaviour: can imply spontaneous life-like development, when it may simply mean unexpected system-level behaviour.
Agency and intention terms
- Agent: implies something that acts with its own agency.
- Autonomous agent: strengthens the impression of independent will.
- Goal-seeking: implies purpose or desire.
- Intent: implies an inner aim or meaning.
- Wants: implies desire.
- Tries: implies effort.
- Attempts: implies intention.
- Chooses: implies free selection or judgement.
- Decides: implies deliberation.
- Prefers: implies taste or subjective ranking.
- Refuses: implies moral or personal resistance, rather than policy-constrained output.
- Cooperates: implies social intention.
- Collaborates: implies partnership with a thinking agent.
- Self-corrects: implies self-awareness and independent error recognition.
- Plans: implies foresight and purpose.
- Strategises: implies deliberate tactical thought.
Emotion and personality terms
- Confidence: often used for probability or fluency, but suggests a psychological feeling of certainty.
- Uncertainty: can suggest doubt rather than a distribution over possible outputs.
- Curiosity: implies intrinsic motivation.
- Surprise: implies expectations and subjective reaction.
- Frustration: implies emotional state.
- Alignment: can sound like moral education or character formation.
- Reward: suggests pleasure or motivation, though it usually means an optimisation signal.
- Preference: can imply liking, especially in “preference learning.”
- Personality: implies a self or character rather than a designed interaction style.
- Helpful, harmless, honest: useful shorthand, but they frame the system as if it has virtues.
Perception terms
- Sees: implies visual experience rather than image processing.
- Looks at: implies directed visual attention.
- Reads: implies human-like reading and comprehension.
- Listens: implies attentive hearing.
- Watches: implies conscious observation.
- Recognises: can imply human perception rather than classification.
- Notices: implies awareness.
- Observes: implies a conscious observer.
- Perceives: strongly implies subjective experience.
- World model: suggests an inner mental model of reality, though this is contested and often metaphorical.
Social and identity terms
- Assistant: implies a social role.
- Copilot: implies a skilled partner.
- Teammate: implies membership and shared purpose.
- Companion: invites emotional attachment.
- Tutor: implies intentional teaching.
- Coach: implies concern for development.
- Researcher: implies professional judgement and responsibility.
- Expert: implies grounded expertise.
- Persona: implies a character or self.
- Roleplay: can imply the model is adopting an identity, rather than producing text in a requested style.
Better neutral replacements
- Hallucination → unsupported output, false generation, fabrication, confabulated output.
- Thinks → generates, outputs, predicts, computes.
- Understands → models linguistic patterns, represents associations, parses input.
- Knows → contains encoded information, has access to retrieved information.
- Remembers → stores, retrieves, includes in context.
- Learns → is trained, is updated, is fine-tuned.
- Decides → selects, ranks, samples, routes.
- Refuses → is blocked by policy, does not produce that output.
- Sees → processes image input.
- Listens → processes audio input or transcribes audio.
- Confidence → probability estimate, score, calibration measure.
- Reward → optimisation signal.
- Goal → objective function, instruction, task specification.
Leave a Reply