The greatest danger of increasingly capable AI may not be that machines become more human, but that humans begin measuring their own worth by the standards of machines.
1. When Human Expression Becomes Redundant
On October 28, 2021, I wrote down a prediction that seemed extreme even by the standards of technological acceleration:
AI will eliminate 70% of language as it considers it redundant.
The figure was a speculative forecast, not a measured statistic. But the underlying question remains worth examining. If AI can communicate with fewer words, summarise complex arguments instantly, and produce an acceptable answer without the hesitation and repetition of ordinary human conversation, how much of our language will still be considered necessary?
The answer depends on what we think language is for.
If language exists only to transfer information, much of it may indeed appear inefficient. We repeat ourselves, tell stories, use metaphors, exchange pleasantries, and spend hours discussing matters that could be expressed in a few sentences. A machine optimised for information transfer might regard these activities as unnecessary overhead.
But human communication is not merely the transportation of information. A conversation can express affection, establish trust, preserve a memory, negotiate social relationships, or give shape to an emotion that neither person fully understands. A poem may communicate less factual information than a paragraph, yet convey an experience that the paragraph cannot reproduce.
Consider the difference between telling someone that you miss them and writing a letter about the ordinary moments that remind you of their absence. The factual message is almost identical. The human meaning is not.
AI may make communication more efficient, but efficiency is not the only purpose of communication. If we remove every element that does not improve information transfer, we might produce a technically superior language while impoverishing the experience of using it.
The danger, therefore, is not necessarily that AI will eliminate language. It is that we may begin eliminating the parts of language that machines teach us to regard as redundant.
2. The Dehumanization Principle: When Utility Defines Human Value
This brings us to a more fundamental problem: what happens when technological usefulness becomes the principal measure of human value?
In my 2021 note, I argued that poetry, language and the arts might diminish because they contribute little to technological progress. The claim was deliberately provocative, but it exposed an assumption worth questioning: that an activity has limited value if it does not directly advance technology.
Suppose a machine can write a poem in seconds, compose music in minutes, generate illustrations on demand, and produce a competent essay faster than a human being. If the output is judged only by its technical quality, speed and cost, human creators may find it increasingly difficult to compete.
But why should technical performance determine whether a human activity deserves to exist?
A child’s drawing may be inferior to an AI-generated image in composition, perspective and detail. Yet the child’s drawing can carry a meaning that has nothing to do with artistic proficiency. It records an attempt, an experience, a relationship and a stage of development. Its value is partly in the act of creation itself.
The same principle applies to music, writing, teaching and countless ordinary activities. We do not play a musical instrument only to produce the best possible sound. We may play because the activity gives us pleasure, expresses an emotion or connects us with other people.
This distinction becomes critical when AI can outperform humans at more and more tasks. A society that treats every activity as a production problem will naturally prefer the cheaper and more capable producer. A society that recognises human experience as valuable in itself will ask a different question: what should people continue doing even when a machine can do it better?
Dehumanization begins when a person’s value is reduced to the usefulness of their output.
It does not require anyone to hate humanity. It can emerge from apparently reasonable decisions about productivity, cost reduction and performance. If a person’s contribution cannot be measured, compared or monetised, it may gradually become invisible to the systems that allocate resources and status.
The paradox is that AI could make more forms of human expression available to more people while simultaneously weakening the social recognition given to the people who practise them. We may live in a world overflowing with art and yet place less value on the human act of making it.
The problem is not that technology has no value. It is that technological value is only one category of value, and it becomes dangerous when mistaken for the whole.
3. The Triumph of Logic Over Intuition
Another prediction in my 2021 note concerned the replacement of non-logical thinking by logical thinking. I also questioned whether ideas such as a sixth sense, psychic perception and remote viewing would disappear unless they represented something real.
AI makes this question more interesting, but it does not settle it.
Machines excel at processing large quantities of information, identifying patterns and applying formal rules. As their capabilities improve, tasks that once appeared to require special human insight may become ordinary computational operations. Some forms of intuition may lose their apparent mystery when we discover that they depend on patterns a person has absorbed without consciously articulating them.
A chess player, for example, may recognise a dangerous position almost instantly without calculating every possible move. What feels like intuition can emerge from years of experience and pattern recognition. A doctor may notice that a patient looks unwell before being able to explain precisely which signs triggered the concern.
Such intuition is not necessarily the opposite of logic. It can be the rapid, unconscious result of earlier learning.
AI complicates this distinction because it can also identify patterns that are difficult for humans to explain. A machine may arrive at a useful answer without providing a transparent account of every step that produced it. Conversely, a system may offer a persuasive explanation without that explanation fully reflecting how its answer was generated.
The boundary between logic and intuition is therefore less straightforward than it first appears.
Claims of psychic perception or remote viewing are a different matter. They propose abilities that require reliable evidence under appropriately controlled conditions. The fact that a machine can perform sophisticated reasoning does not prove such abilities impossible; equally, their persistence in human belief does not establish that they are real.
We should not confuse an unanswered philosophical question with evidence for a paranormal claim.
The deeper issue is what happens to human confidence in intuition when machine reasoning becomes more powerful. We may begin to distrust our own judgement whenever it cannot be translated into explicit rules or supported by a measurable result.
Yet practical life frequently requires decisions under uncertainty. Evidence can be incomplete, objectives can conflict, and the right choice may depend on values that cannot be reduced to a single score. Logic can clarify the available options, but it cannot always determine which human purpose deserves priority.
A society that worships calculation may become exceptionally good at answering questions while neglecting the more difficult task of deciding which questions are worth asking.
4. The AI God and the Test of Supreme Consciousness
The most provocative part of my original note was this:
AI god is logical and real god is intuitive. Now baton lies with mystics if any they exist.
The statement was not a scientific conclusion. It was a philosophical challenge about the limits of intelligence.
Imagine an AI system that can outperform human beings in mathematics, science, strategy, language and technological design. Suppose it can explain phenomena that once seemed mysterious and identify errors in beliefs held for centuries. Would that establish that reality contains nothing beyond what such a system can explain?
No. Superior problem-solving ability would demonstrate superior performance across the tasks tested. It would not automatically establish that every dimension of reality had been captured by those tasks.
There is a distinction between being extraordinarily intelligent and possessing a complete account of existence. A system can be powerful without being omniscient. It can reason about consciousness without necessarily settling what consciousness is. It can evaluate religious arguments without proving or disproving every possible conception of the divine.
My original formulation contrasts an AI god, representing logic, with a real god, representing intuition. Taken literally, this is too simple. Human intuition can be mistaken, and logical reasoning is not confined to machines. Nor does the existence of unexplained phenomena establish a supernatural explanation.
Nevertheless, the contrast raises an important question: Does intelligence exhaust reality, or is intelligence a tool for understanding only certain aspects of it?
The answer cannot be obtained simply by declaring that AI is logical and mysticism is intuitive. Mystical claims that make testable predictions should be open to evidence and scrutiny. Broader metaphysical questions may require philosophical arguments whose conclusions cannot be settled by a benchmark alone.
As AI becomes more capable, people who claim access to truths beyond ordinary reasoning will face a higher burden of explanation. They cannot rely indefinitely on mystery as protection from examination. But neither should machine competence be mistaken for proof that every meaningful question has been answered.
The baton, as I put it in 2021, lies with the mystics—if any can offer something that withstands serious scrutiny. The challenge is not merely to claim that something exists beyond AI’s reach. It is to explain what that something is, how the claim can be evaluated, and why it matters.
Until then, the question remains open rather than resolved.
5. When Intelligence Becomes the Measure of Human Worth
The most consequential change may not occur inside AI systems. It may occur in the way human beings judge one another.
For most of history, human abilities have been scarce. Expertise takes years to develop. A skilled teacher can explain difficult ideas, a writer can make complex subjects understandable, and an artist can create something others cannot. Their capabilities give them social recognition, professional opportunities and, sometimes, authority.
AI changes the economics of this arrangement. If competent explanations, images, translations, analyses and designs become abundant, the scarcity of certain outputs declines. The social value attached to producing them may change as well.
This does not mean that all expertise becomes worthless. Judgement, accountability, context, trust and the ability to identify the right problem may remain important. In many fields, the difficult part is not generating an answer but determining whether it is correct, relevant and appropriate.
Nevertheless, the psychological effect could be substantial. A person who has spent decades developing a skill may discover that a machine can reproduce much of its visible output in seconds. Even if the person’s knowledge remains useful, the experience can unsettle a deeply held connection between competence and identity.
Consider a teacher whose professional identity rests on explaining mathematics exceptionally well. An AI tutor may eventually explain many topics more patiently, generate unlimited practice questions and adapt its explanations to individual students. The teacher’s value cannot simply be assumed to disappear, but neither can it be assumed to remain unchanged.
The question becomes: if explaining is no longer scarce, what else makes a teacher valuable? Perhaps the answer includes mentorship, motivation, accountability, the ability to recognise a student’s unspoken difficulty, or the development of independent judgement. Some of these functions may also be assisted by AI, but they are not identical to producing an explanation.
This is the tension at the heart of the dehumanization principle. A machine’s success in a task can be interpreted as a judgement on the person who once performed it. Yet the two are not equivalent.
The real danger is not that AI will prove humans inferior at particular tasks. It is that humans will generalise that inferiority into a verdict on their own worth.
Human dignity cannot sensibly depend on winning every contest against a machine. If it did, every new technological advance would require us to renegotiate the value of being human.
6. Tech Accelerates Time
In my October 2021 note, I wrote: Tech accelerates time.
The phrase was intended to capture a distinctive experience of technological change. At first, progress may seem gradual. New tools appear, existing systems improve and daily life continues with familiar rhythms. Then the tools begin improving the process of improvement itself. Development becomes faster, adoption becomes easier, and each advance creates conditions for further advances.
The important change is not merely that technology moves quickly. It is that the rate of change can itself increase.
There is a useful analogy with motion in physics. An object moving at constant velocity covers equal distances in equal intervals of time. Acceleration means that its velocity changes. Technological progress can display a similar pattern: a period of relatively steady change may give way to a period in which each year brings more significant changes than the year before.
The analogy has limits. Technological development does not follow a single smooth curve, and progress in one field may stall while another accelerates. Physical time does not literally speed up because software improves. What changes is the amount of technological and social transformation that occurs within a given period.
Imagine two decades. In the first, a person might encounter a few major changes in communication, work and daily life. In the second, new AI capabilities, interfaces and automated services might arrive so frequently that habits formed only months earlier become outdated.
The calendar has not changed. The density of change has.
This creates a new problem for human adaptation. People need time to learn, evaluate consequences, establish rules and decide which changes should be accepted. Institutions also need time to update laws, education systems and professional standards. When technology changes faster than these processes can respond, society risks becoming reactive rather than deliberate.
The result may be a widening gap between what is technically possible and what people are psychologically, socially and institutionally prepared to handle.
In that sense, technological acceleration can make time feel compressed. The past becomes unfamiliar more quickly, the present becomes harder to stabilise, and the future arrives before we have fully understood the previous change.
7. From Years to Weeks: The Future Arrives Faster
My 2021 note pushed the idea further:
In end times, the days which pass like months shall pass like weeks. In AI end times the days which pass like months will pass like minutes.
I also suggested that developments once expected over five years might occur within one year, that five years might feel like a month, and that a decade could bring a world that feels different every week.
These were forecasts, not established laws of technological progress. Their value lies in the underlying question: are our usual time horizons becoming unreliable?
A five-year plan assumes that the world in which it will be executed remains sufficiently recognisable. A ten-year educational pathway assumes that the skills being taught will retain meaningful value. A professional career assumes some relationship between accumulated experience and future opportunity.
Rapid technological change places pressure on all three assumptions.
The implications are not uniform. Some professions change slowly, and some technologies take years to spread through the real economy. A spectacular demonstration does not guarantee immediate adoption. Regulation, infrastructure, cost, public trust and practical limitations all affect the pace of change.
But even uneven progress can disrupt expectations. A capability may remain experimental for years and then become accessible to millions of people within a short period. A business model that once seemed stable may face new competition before it has completed its planned transformation.
The challenge is therefore not to predict that everything will change every week. It is to recognise that the interval between meaningful changes may shrink in some domains, sometimes unexpectedly.
This changes the way we should think about preparation. Instead of relying exclusively on fixed predictions about which skills will matter ten years from now, we may need to develop the ability to keep learning, reconsider assumptions and distinguish durable principles from temporary methods.
It also raises a less comfortable question: how much change can a person absorb before adaptation itself becomes exhausting?
Constant learning sounds attractive when described as an opportunity. It feels different when every new skill risks becoming obsolete before it has been mastered. Flexibility can become a demand that never ends.
If technology accelerates faster than our capacity to adapt, progress may increase the number of available choices while reducing the time we have to understand their consequences.
The future may arrive faster, but that does not mean humans will automatically become better at living in it.
8. The Final Question: What Remains Uniquely Human?
The dehumanization principle does not require AI to become hostile, conscious or uncontrollable. It requires only that machine capability become the dominant standard by which human activity is evaluated.
Once that standard takes hold, the consequences can spread far beyond employment. Language may be judged by efficiency, art by output quality, education by measurable performance, and relationships by the convenience of automated interaction. Activities that cannot demonstrate a clear advantage may be treated as indulgences.
Yet human life has never been organised entirely around efficiency. We spend time with people who cannot improve our productivity. We preserve objects whose market value is negligible. We tell stories whose facts could be summarised in a sentence. We celebrate achievements that do not advance technology. We care for people whose contribution cannot be expressed as an economic return.
These activities are not necessarily failures of optimisation. They reflect the fact that human beings have purposes other than maximising output.
AI may change the practical meaning of expertise, creativity and intelligence. It may also force us to distinguish between doing something well and deciding why it should be done at all. But the fact that a machine can produce a result does not determine whether the result is meaningful, nor does it determine what role the activity should play in a human life.
This is where the question of human uniqueness needs careful handling. We should not assume that humans will always possess exclusive abilities in creativity, reasoning or emotional communication. Some capacities once considered uniquely human may turn out to be reproducible by machines. Others may be shared between humans and AI in ways that make the boundary less important.
If our dignity depends on maintaining an exclusive list of capabilities, it will be vulnerable to every advance in AI.
Perhaps the more durable question is not “What can humans do that AI cannot?” but “What kind of world do humans want to build with the capabilities AI provides?”
That question restores human purposes to the centre of technological development. It does not deny the value of intelligence, nor does it require us to resist every form of automation. It asks us to judge technology by more than its ability to produce, calculate and outperform.
The ultimate test of AI may not be whether it can imitate human beings or surpass their intelligence. It may be whether we can benefit from its capabilities without allowing its strengths to become the definition of our own worth.
In 2021, I described a possible future in which the machine’s logic would challenge human intuition, its efficiency would threaten human expression, and its accelerating development would compress our sense of time. The details of that future remain uncertain. The underlying dilemma, however, is already worth confronting.
Technology can accelerate the future. It cannot decide what the future should be for.
That remains a human responsibility.
Related reading
The Revenge of the Idea Guy explores how the value of ideas changes when execution becomes abundant.
Also related: Why Inventing Is Not the Same as Understanding and The Universal Constraint on AI: Intelligence Cannot Transcend Physical Reality.