Part 2 of the AI Productivity Series
When machines eliminate scarcity, productivity may stop being the problem. Human purpose, agency and meaningful roles may become the new bottlenecks.
Previous: Part 1 — Are AI Productivity Gains Worth It?
1. From Productivity to Abundance
Part 1, Are AI Productivity Gains Worth It?, asked whether AI productivity gains are actually worth as much as they appear. The answer was not that AI productivity is unimportant. It was that productivity does not exist in isolation. AI can accelerate cognition, but the real world still has limits: infrastructure, physical resources, human attention, human time, perception, consumption and demand.
Part 2 pushes that argument one step further. Suppose those constraints are progressively removed. Suppose AI becomes capable of designing, planning and coordinating almost everything, while robotics can build, move, grow, repair and deliver almost everything humans need.
At that point, the central question changes. The problem is no longer simply how much we can produce. It becomes: what happens when production itself stops being the main constraint?
AI may eventually create a strange situation in which capability becomes extraordinarily abundant while the human capacity to use that capability remains stubbornly finite.
2. The Infinite Production Problem
Imagine an AI system that can generate a new novel in seconds, a film in minutes, a game in hours, a scientific hypothesis on demand, or a completely new virtual world whenever someone asks for one. Add robotics, and the same principle extends into the physical world: food can be grown, houses can be constructed, goods can be manufactured and services can be delivered with progressively less human labour.
The machine’s production capacity can approach infinity while human consumption remains stubbornly finite.
We cannot eat a thousand meals simply because food becomes almost free. We cannot watch ten thousand movies in a day. We cannot read a million books. We cannot live in a thousand houses at the same time.
The important point is not that abundance is bad. Abundance is enormously valuable when it removes genuine scarcity. The question is what happens after the scarcity has already been removed.
3. Capability Is Not Utility
One of the easiest mistakes in thinking about powerful AI is to confuse what a machine can generate with what a human can actually use.
Consider chess. A system such as Stockfish can analyse enormous numbers of positions and variations. Its computational space can be vastly larger than the experience available to one human player. Yet a person sitting at a chessboard still plays one game at a time.
AI can therefore have enormous capability without creating an equal amount of human experience. The distinction becomes even more important when AI begins generating books, films, strategies, designs, images and ideas at industrial scale.
Capability ≠ Production ≠ Utility ≠ Consumption. These are different stages of the chain, and the further AI moves along it, the more important the distinction becomes.
4. The Human Attention Ceiling
The first scarce resource after abundance may be attention.
A sufficiently powerful AI could produce more things than anyone could possibly inspect. Imagine millions of excellent books being created every day, thousands of new films appearing every hour, and an effectively unlimited supply of games, courses, music, research papers and simulated experiences.
Human beings still have only 24 hours in a day. Even with better recommendation systems, we cannot consume everything that exists. A machine can produce ten thousand possibilities while a human may have time to examine only ten.
This creates a peculiar inversion. In the old economy, producers competed because supply was scarce. In an AI-abundant economy, the competition may increasingly be for the scarce attention of the consumer.
5. Even Perception Has a Ceiling
The limit is not only time. Human perception itself has boundaries.
An AI can generate an image at increasingly high resolution, reconstruct details invisible to the unaided eye, or create increasingly precise simulations. But beyond some point, additional machine capability may stop creating a meaningful difference in human experience.
The same principle appears in sound, visual detail, simulation quality and information density. More measurable quality does not automatically mean more experienced quality.
This is another form of diminishing return: the capability curve can keep rising while the human experience curve begins to converge.
6. The Bank That Can Process Everything
Consider a bank. AI could make transactions dramatically faster. Systems could process enormous numbers of payments, verify information almost instantly and automate much of the surrounding workflow.
But faster processing does not create customers who do not exist. A bank may be able to process transactions one hundred times faster, yet the economy does not automatically contain one hundred times as many genuine transactions.
The same pattern appears everywhere. Faster production is not the same as greater demand. Greater supply is not the same as greater consumption. And greater capability is not the same as greater human welfare.
The bottleneck simply moves.
7. When Scarcity Disappears
Now take the argument to its logical extreme. Imagine robots that can grow food, construct houses, maintain infrastructure, manufacture goods and provide many routine services with minimal human intervention.
If the material requirements of life become extremely cheap and abundant, the old relationship between work and survival begins to weaken.
For most of human history, people had to work because the world imposed requirements on them. Food had to be produced. Shelter had to be built. Goods had to be transported. Knowledge had to be taught. Machines could not simply do all of it.
What happens when machines can?
8. What Happens When Humans Are No Longer Needed?
This question sounds alarming because modern society is organised around necessity. Jobs do not merely provide income. They also provide structure, identity, status, routine, social contact and a sense that one’s effort has a purpose.
If machines perform most economically necessary tasks, the disappearance of scarcity does not automatically create a disappearance of human need. It creates a different problem: how should human life be organised when survival no longer supplies the structure?
A society can be materially successful and still face questions about what its members are supposed to do with their time, attention and freedom.
The deeper transition is therefore not simply from human labour to machine labour. It is from necessity-driven activity to choice-driven activity.
9. The Agency Problem
This is where abundance can become surprisingly difficult. A system that can optimise almost every aspect of life might be able to recommend the best career, the best diet, the best partner, the best schedule, the best entertainment and even the best sequence of experiences.
Yet a perfectly optimised life is not necessarily the same thing as an authored life.
There is a difference between being helped to choose and having the choosing done for you. The uncomfortable human response may be: “Don’t optimise my life. Let me screw it up myself.”
Agency means retaining the ability to choose, experiment, fail, change direction and sometimes deliberately reject the optimal path. In an abundant world, preserving that freedom may become more important, not less.
10. The Problem of Infinite Stimulation
AI could also make entertainment effectively unlimited. Every person could have personalised novels, films, games, music, conversations and virtual worlds generated on demand.
At first this sounds like paradise. But repeated exposure to rewards can reduce their novelty. Human satisfaction is not simply a function of how much stimulation is available. Familiarity, expectation, comparison and adaptation all influence how an experience is valued.
If every fantasy can be generated instantly, the scarcity of entertainment may disappear without creating an infinite capacity to enjoy it.
The problem may shift from “How do we get something interesting?” to “How do we decide what is worth experiencing?”
11. AI Can Remove the Limits of the World Faster Than the Limits of the Mind
This may be the central paradox of the entire transition.
AI and robotics can progressively remove external constraints. They can make information abundant, production cheap, services available and physical tasks increasingly automated. But the human mind still operates through a finite body, finite attention and finite time.
AI may therefore remove the limits of the external world faster than it removes the limits of the human mind.
That creates an unusual asymmetry: the world of possibilities can expand much faster than the space of possibilities that one person can actually explore.
12. The Calhoun Question
The famous Universe 25 experiments by John B. Calhoun are sometimes used as a warning about what happens when animals are placed in conditions of abundance. The popular story focuses on unlimited food, protection from predators and the eventual breakdown of social behaviour.
But this should not be treated as evidence that humans will inevitably collapse when material scarcity disappears. The experiment involved mice in highly artificial conditions, and interpretations of what it demonstrated have been debated.
The more useful question for this article is narrower: what happens when a society loses many of the constraints that previously forced individuals to occupy meaningful roles?
Calhoun later discussed ideas such as “conceptual space” and the ability of populations to generate new patterns of social behaviour. That points toward a more interesting possibility: when old roles disappear, the challenge may be to create new ones rather than simply assume that nothing will replace them.
13. From Jobs to Meaningful Roles
For most of history, human roles were built around necessity. Farmers produced food. Builders produced shelter. Teachers transmitted knowledge. Doctors treated illness. Engineers built infrastructure. Workers manufactured goods.
AI and robotics could eventually automate large parts of these functions. That does not mean humans become useless. It means the reason for doing the work may change.
Humans may create roles around exploration, research, art, competition, community, discovery, teaching, relationships and experiences that are valuable precisely because people choose them.
The question becomes: what happens to a society when machines don’t merely take its jobs, but gradually take away the reasons those jobs existed?
14. The New Scarcity
When production becomes abundant, scarcity does not necessarily disappear. It moves.
First, scarcity moves from goods toward attention. Then it can move from attention toward meaning. Finally, it may move toward agency: the ability to decide what matters and what to do with one’s finite life.
The scarce resource becomes the human being’s capacity to experience, choose and care.
Machines optimise outcomes. Humans often value experiences. That distinction may become one of the defining economic and philosophical differences of an AI-abundant civilisation.
15. The Ultimate Productivity Problem
At the beginning of the industrial era, the problem was producing enough. Later, the problem became producing efficiently. AI pushes the next stage toward producing almost anything on demand.
But there is a point at which production itself stops being the central problem.
If I can create a thousand books, a hundred films, millions of images and countless ideas, the next question is not whether I can make more. It is who will use them, who will experience them and why they should exist.
The ultimate productivity problem may not be that AI cannot produce enough. It may be that AI can produce far more than humanity has any practical use for.
16. What Will Humans Choose to Do?
This question has no single technological answer because it is ultimately a human choice.
Some people may explore. Some may create. Some may compete. Some may research. Some may build communities. Some may teach even when teaching is no longer economically necessary. Others may simply enjoy relationships, nature, games, travel, craft or contemplation.
The important change is that these activities could increasingly be chosen rather than imposed by economic necessity.
The future may therefore require a new social architecture: not a system that tells humans what they must do to survive, but one that helps them discover what they genuinely want to do when survival is no longer the organising principle.
17. The Human Bottleneck
This brings the argument back to productivity.
AI can become extraordinarily productive. Robotics can become extraordinarily capable. Energy, manufacturing and computation can all improve. Yet the human being remains bounded by time, attention, perception, embodiment and the need to experience life sequentially.
The bottleneck therefore keeps moving. First it is computation. Then production. Then distribution. Then demand. Then attention. Then meaning. Eventually the bottleneck may be the human capacity to decide what deserves to exist at all.
The future challenge may therefore be less about teaching machines to produce more and more about teaching ourselves how to choose among an almost unlimited set of possibilities.
18. The Question After Productivity
Part 1 asked whether AI productivity gains are worth it when the physical and human world imposes ceilings on their usefulness.
Part 2 takes the next step. Suppose AI eventually pushes many of those ceilings outward. Suppose production becomes abundant and machines can perform most necessary work.
Then the great question changes.
It is no longer simply: “What can AI produce?” It becomes: “What is worth producing?” And after that comes the deeper question: “When machines can produce almost everything we need, what will humans choose to do simply because it is worth doing?”
That may be the question that follows productivity itself.
Takeaway
AI can make capability abundant without making human experience unlimited. A machine may generate millions of chess variations, books, films, images, strategies and simulated worlds, but a human still has one mind, one body, 24 hours a day and one life. As AI and robotics reduce material scarcity, the bottleneck may move from production to attention, then from attention to agency and meaning. The assumption that a more productive civilisation must automatically be a more fulfilled civilisation therefore deserves reconsideration. The central question changes from “What can AI produce?” to “What is worth producing?” And eventually: when machines can produce almost everything we need, what will humans choose to do simply because it is worth doing?
The Cute Tiger Cub Problem