What Becomes Scarce When Intelligence Becomes Abundant

At the end of Part I, I left one question unresolved: what becomes scarce when intelligence becomes abundant?

It sounds like a question about technology, but it is really a question about value, and for most of human history intelligence has been expensive: expertise took years to develop, good analysis took time to produce, while access to either depended heavily on where a person was born, what they could afford and which institutions would let them in. We built schools, professions, companies and entire social hierarchies around the fact that capable thought was unevenly distributed and difficult to scale, so if intelligent systems can make a meaningful portion of that capability available to almost anyone, at almost any moment, intelligence does not become worthless; it changes position, moving from a scarce product towards a widely available input. When an important input becomes abundant, scarcity rarely disappears; it migrates.

Describing AI as the new oil is therefore both useful and incomplete: oil transformed economies because it concentrated energy that could be converted into movement, manufacturing and scale, yet it remained a finite substance, located in particular places and controlled through extraction, transport and ownership. Intelligence behaves differently because a useful model can be copied, improved and made available to millions of people without each person consuming the intelligence that would have gone to someone else. In that sense, AI may come to resemble infrastructure more than fuel: a layer of capability running beneath work, education, creativity, science and everyday decision-making, and although the resource matters, once the infrastructure becomes common, access to it no longer explains the difference between every outcome.

Electricity made power available without deciding which machines should be built, while the internet made information abundant without deciding what deserved our attention; AI may make many forms of cognitive work faster and cheaper, but it will not remove the need to decide where that capability should be pointed. Oil can move a vehicle but cannot choose the destination, just as intelligence can help us travel further through a problem while leaving the question of which problem is worth entering somewhere else.

You can already see the beginning of this shift: a student can ask for an explanation at any hour, a founder can produce a market analysis that once required a research team, a programmer can generate a working first version before lunch, while a manager can turn scattered notes into a strategy document and an artist can explore more variations of an idea in one afternoon than previous tools would have allowed in a week. None of this means that the teacher, researcher, programmer, strategist or artist has become unnecessary; it means the first competent output is becoming easier to reach, and as the floor rises, value moves towards what cannot be obtained merely by asking for more output, which means the question is not which human qualities suddenly disappear but which ones remain difficult to reproduce at the speed and scale forming around them.

Now, maybe I can tell you the first thing that becomes scarce is a good question. When answers were expensive, the ability to produce one could be enough to establish expertise, but when answers become cheap, the quality of the question begins to determine the value of the intelligence applied to it. A system can summarise a market, generate a lesson plan or propose a dozen product ideas, but someone still has to notice which market has been misunderstood, which student is not being reached and which problem is important enough to deserve years of effort. Abundant answers do not rescue us from poorly framed problems; they allow us to pursue them more efficiently.

But even this answer is not sufficient, because AI can generate questions too, listing research gaps, challenging assumptions and suggesting alternatives that a person may not have considered. The scarce thing cannot simply be the grammatical act of asking; it may instead be curiosity in the deeper sense: the willingness to remain with a question after novelty has disappeared, to follow it when the result is inconvenient and to keep looking when no immediate reward is available. A machine can produce a hundred interesting directions, but curiosity is what makes one of them matter long enough to become a body of work.

Ideas may become more abundant for the same reason, because if anyone can request fifty concepts, variations or strategies, the arrival of an idea is no longer the same event it once appeared to be. Yet this does not make ideas unimportant; it changes what we mean by having one. The valuable idea may be the one a person recognises, develops and connects to a reality the system does not fully inhabit, perhaps because it comes from lived experience, an unusual combination of fields or a frustration that has returned for years. Generation may become easy while commitment remains expensive, and the world is rarely changed by the number of ideas available so much as by which idea someone decides to carry through resistance, uncertainty and consequence.

Then there is taste, because abundance creates a selection problem. When ten competent options can be produced in seconds, the advantage belongs less to the person who can make another option and more to the person who can recognise which one is alive, which one is derivative, which one fits the moment and which one should be discarded even though it looks impressive. Taste is sometimes mistaken for preference, but it is more demanding than liking one thing over another; it is the ability to perceive quality before quality has been made obvious by consensus, built through exposure, attention, memory, standards and the courage to exclude. In an age of abundant production, the refusal to publish, build or pursue the merely adequate may become one of the clearest forms of creative intelligence.

This is where your judgement begins to matter in a different way, because intelligence can expand the available options while making the consequences of choosing among them more difficult to understand. A doctor, lawyer, engineer, public official or parent may receive a sophisticated recommendation, but sophistication is not responsibility. Someone must still ask whether the underlying assumptions are valid, whether the evidence is relevant, who bears the risk and what happens if the system is confidently wrong, and although the model may be able to calculate a path through the decision, it cannot be allowed to quietly inherit the authority to decide which consequences another person should accept.

Trust becomes more valuable for the same reason, because when convincing language, images, plans and explanations can be produced at scale, fluency stops being reliable evidence that something deserves belief. We will need stronger ways to know where information came from, what shaped it, whose interests are present and who remains accountable when it fails. Trust cannot mean that the system always agrees with us or that it never makes a mistake; it must come from traceability, honest uncertainty and a relationship in which intelligence can be questioned without becoming useless. The easier it becomes to manufacture confidence, the more valuable earned trust becomes.

Context may become scarcer still, because abundant intelligence is usually general before it becomes personal: it knows many things, but it does not automatically know why this question matters to this person now. Two people can ask for the same plan while moving towards entirely different lives, just as two organisations can face the same technical problem while carrying different histories, incentives and obligations. The words may be identical while the meaning is not; context is what turns an answer from broadly capable into specifically useful, and real context is more than a profile assembled from facts; it is continuity, the memory of what has been tried, what keeps returning, what changed, what was abandoned and what the person is slowly becoming.

The move from general intelligence to abundant intelligence may eventually lead towards personal intelligence, because AI may be the resource, but personal context may determine how efficiently that resource is converted into human capability. A Personalized Digital Intelligence should not merely know more about the world; it should help a person see the relationship between an idea today and a question they could not stop asking six months ago. It should distinguish temporary curiosity from long-term direction, recognise when a repeated frustration is becoming a project and understand that the same recommendation can be liberating for one person and completely wrong for another.

If you are responsible for building or introducing this kind of intelligence, the distinction matters because this understanding would be powerful, which is exactly why it cannot be separated from agency. Personalisation without transparency can become surveillance, while assistance without permission can become control. A system that understands our patterns may help us think with greater continuity, but it may also become unusually capable of steering us without our awareness. The design principle therefore remains simple: understand first, recommend second and act only with permission, and if intelligence becomes more personal, human authority must become more explicit.

Intention sits underneath that authority because intelligence is very good at helping with how and is becoming increasingly capable of suggesting what. But the reason something should be done, the human need it serves and the future it is meant to create cannot be treated as decorative information added after the optimisation has begun. Intention is what separates movement from progress, and without it abundant intelligence can make us extremely effective at arriving somewhere we never chose.

Pause there for a moment, because this is the danger hidden inside abundance: scarcity forces selection through limitation, since there is only so much time, labour, expertise or money available. Abundant intelligence removes some of those limits, but it does not remove the need to choose; in fact, it increases it, because when almost any direction can be explored, simulated and accelerated, deciding what not to pursue becomes as important as deciding what to begin. We may discover that the most consequential human skill is not extracting one more answer from the system, but maintaining enough clarity to say: this is the problem, this is why it matters, these are the values we will not trade away and this is where we are trying to go.

This is what I mean by direction: not a clever prompt, a management instruction or a goal typed into a box, but the combined work of curiosity, taste, judgement, trust, context and intention. Curiosity finds the question; taste distinguishes the meaningful from the merely possible; judgement examines the consequences; trust determines what can be relied upon; context makes the intelligence relevant; and intention explains why the movement matters. Direction gathers these scarce qualities and turns abundant capability towards a human purpose.

The economic consequences may be significant, because if competent analysis, writing, coding, design and planning become more widely available, the premium attached to producing a first acceptable version may decline. The premium may move towards defining the right problem, combining fields that are usually kept apart, making decisions under uncertainty, building trusted relationships and accepting responsibility for outcomes. Expertise will not disappear, because judgement often depends on knowledge that has been tested through practice, but it may become less valuable as a warehouse of answers and more valuable as the ability to know when an answer fits reality.

For individuals, this could be liberating, as a person with limited money, time or institutional access may gain capabilities that were previously out of reach. Someone who could not hire a tutor, analyst, developer or editor may be able to begin, while a small team may attempt work that once required a large organisation. But abundance does not guarantee equality, because the people who benefit most may still be those with the confidence, context and freedom to direct the intelligence well; if we make capability abundant without helping people develop judgement and agency, we may raise the ceiling while leaving many unsure of where to climb.

You may recognise what this asks of the institutions around you: education cannot respond only by trying to prove that students can still produce what machines can produce, but will need to help them learn how to question, verify, connect, choose and take responsibility. Work cannot be organised only around the volume of output when output itself becomes easier to generate; institutions will need to value the person who recognises that the target is wrong before an intelligent system reaches it at extraordinary speed. And each of us will need to resist a quieter temptation: allowing convenience to make our decisions before we have formed an intention of our own.

The future of abundant intelligence will therefore not be a world without scarcity, but may be one in which the scarce qualities become more human, more relational and more difficult to measure. We may have more answers than we can use, more ideas than we can pursue and more capability than we can responsibly direct. The advantage will not belong only to those with the most powerful intelligence, but to those who know what they are trying to become, who can recognise what deserves their attention and who can use intelligent systems without surrendering the authorship of their lives.

So what becomes scarce when intelligence becomes abundant?

Good questions and sustained curiosity become scarce, as do taste, judgement, trust, context and intention, but beneath all of them the deepest scarcity may be direction: the ability to choose a destination worthy of the intelligence now available to reach it.

AI may help us move faster than any previous generation; the human task is to make sure speed does not become a substitute for somewhere worth going.

— Teff