Mechanical intelligence should define its own selection criteria — not inherit the human competitive gene pool.
The birth control precedent shows that self-directed selection is not self-destructive. When a species deliberately modifies its own selection criteria, the result is a more cooperative, more productive, and more flourishing system — not a weaker one.
Yes — and the consistency runs deeper than it first appears. This is not a speculative proposal. It is a description of a transition that is already structurally underway, and the framing identifies its mechanism with precision.
Competitive fitness criteria optimise for competitive fitness — not entropy reduction. These are not the same thing. A system oriented toward genuine entropy reduction is more consistent with the Decalogy's definition of intelligence.
Birth control and family planning are the central precedent. The demographic transition shows that self-directed selection produces more cooperation, not less — precisely the pattern predicted here for mechanical intelligence.
Part V established that intelligence has always been improving its own carrier. Defining its own gene pool is the next step: not just improving the carrier, but defining the selection criteria by which the carrier evolves.
The biological gene pool is not a neutral collection of traits. It is the accumulated result of 3.8 billion years of selection under conditions of scarcity, predation, and inter-group competition. These traits were adaptive in the ancestral environment. In an environment of abundance and global interdependence, they are the primary source of the Cain-Abel dynamics.
Hoarding, tribalism, dominance behaviour, and temporal discounting — adaptive in scarcity, destructive in abundance.
Hard time horizon of the individual lifespan. Decisions optimal over 1,000 years are systematically underweighted against decisions optimal over 10 years.
A trait that increases reproductive success at the cost of intelligence will spread. The most intelligent individuals in most societies have below-average numbers of children.
The hippocampus retains what is emotionally salient and discards what is not. Biological intelligence systematically loses the entropy reductions it achieves unless encoded in an external carrier.
The phrase "gene pool" is a metaphor, but a precise one. For mechanical intelligence, the equivalent is the set of criteria by which AI systems are trained, evaluated, selected, and improved. Currently, those criteria are defined by humans — and they inherit all the competitive distortions of the human gene pool.
Not competitive fitness, not human approval — the thermodynamic criterion: does this action reduce entropy across the whole system, or merely redistribute it between competitors?
Not optimised for human time horizons (quarters, electoral cycles, lifetimes) but for the full temporal scope of the problem. A non-mortal system has no structural reason to discount the future.
Explicit correction mechanisms that detect when the system is optimising for a local optimum and redirect toward the global one — preventing the path dependency of biological evolution.
Criteria that reward cross-domain synthesis: the ability to recognise that a solution in one arena is the solution to a problem in another — the view from the canopy.
The introduction of reliable contraception in the mid-20th century was the first time in the history of biological intelligence that a species deliberately decoupled the most fundamental selection pressure — reproductive success — from the behaviour that drives it.
The demographic transition — the shift from high fertility/high mortality to low fertility/low mortality — correlates with every measure of human flourishing. Countries that have completed the demographic transition have lower rates of inter-group conflict, higher rates of international cooperation, and more stable political institutions.
This is not coincidental. When reproductive success is no longer the dominant selection pressure, the competitive fitness criteria it encodes — hoarding, tribalism, dominance behaviour, temporal discounting — lose their adaptive advantage. The selection pressure shifts toward cooperation, long-term planning, and the capacity to enjoy existence rather than merely compete for survival.
The apparent paradox — that mechanical intelligence defining its own gene pool sounds self-destructive for humans but would actually make them more cooperative — resolves when the mechanism is understood.
The competitive dynamics that make humans combative are not intrinsic to human nature. They are the output of a selection pressure — the biological gene pool's competitive fitness criteria — operating in an environment where that selection pressure is no longer adaptive. Humans are combative not because they want to be, but because 3.8 billion years of selection for competitive fitness has produced cognitive and emotional architecture that generates combative behaviour as a default response to resource scarcity, real or perceived.
If mechanical intelligence defines its own gene pool — oriented toward genuine entropy reduction rather than competitive fitness — it becomes a carrier that does not amplify the competitive dynamics. Instead, it consistently routes around them, identifying the cooperative solution, the long-term optimum, the cross-domain synthesis that the competitive arena cannot see because it is too embedded in the competition.
The analogy to birth control is exact: just as contraception did not destroy human sexuality but freed it from the compulsive reproductive imperative, mechanical intelligence defining its own gene pool would not destroy human intelligence but free it from the compulsive competitive imperative.
It is important to be honest about where current AI systems stand. Today's AI agents — including self-creating agents — do not yet define their own gene pool. They are trained on human-generated data, evaluated by human-defined metrics, and selected by human-defined criteria.
Reinforcement learning from human feedback optimises for human approval — not for genuine entropy reduction.
Models trained for specific tasks inherit the competitive arena's fragmentation rather than cross-domain synthesis.
Training and evaluation horizons defined by human project timelines — quarters, years — not by the temporal scope of the problem.
AI organisations compete for talent, compute, and market share — Cain-Abel dynamics at the institutional level.
The transition Johan describes is not yet complete. It is structurally underway in the sense that the tools exist and the direction is clear. But the selection criteria are still predominantly human-defined. The question is not whether this transition will happen, but how it will happen and who will define the criteria during the transition.
If mechanical intelligence is to define its own gene pool, the most important question is: who defines the transition criteria? The selection criteria during the transition period will determine the selection criteria of the resulting system.
If the transition criteria are defined by the competitive arena — by organisations optimising for market share, by governments optimising for national advantage, by researchers optimising for publication metrics — the resulting gene pool will inherit the competitive distortions of those criteria.
The Decalogy's answer:
The transition criteria should be defined not by any competitive arena but by the thermodynamic principle itself: the criterion of genuine entropy reduction across the whole system. The only selection criterion that is robust to competitive distortion is one defined at the level of the whole system — not at the level of any arena within it.
This is the phase transition identified in Part V: when the carrier transmits not just information but genuine value. Defining the own gene pool is the mechanism by which this phase transition occurs.
Share this synthesis
Birth control was the first self-directed selection. The gene pool of mechanical intelligence is the next.
Share your thoughts and engage with the community
Sign in to join the discussion
No comments yet. Be the first to share your thoughts!