Imagine a flood moving down a river shared by several cities. Sensors report rising water and messages travel almost instantly. Yet the hardest questions remain: who should act first, who should pay, and whose losses should count?
The information can arrive before the cooperation does.
That gap suggests a different way to think about humanity’s next evolution. The next leap may not require a larger brain. It may happen between our brains: in how people, institutions, and AI systems learn to understand a problem and act together.
This is a possible organizational transition, not an established biological one. It might depend less on making each of us smarter than on making our cooperation less fragile.
On this page
- The parts can exist before the system
- Humanity was thinking collectively before computers
- The internet shortened one kind of distance
- AI can help connect understanding, not manufacture agreement
- A new species is a much stronger claim
- Shared memory is not enough without shared correction
- A larger “we” must not erase the individual
- Make room for a civilization-sized pause
The parts can exist before the system#
A sponge has no neurons, yet it is not an uncoordinated bag of cells.
A 2021 study published in Science identified 18 distinct cell types in the freshwater sponge Spongilla lacustris. The researchers found secretory neuroid cells in close contact with digestive choanocytes, which expressed scaffolding and receptor proteins. This communication system used molecular modules also found in the pre- and postsynaptic machinery of other animals.
A 2007 study of a different sponge identified a nearly complete set of homologs of the postsynaptic proteins it investigated. Important receptor components were absent: this was not a complete nervous system awaiting activation.
Nor were these proteins necessarily idle. Molecular machinery can serve one function and later be recruited into another. Evolution can reorganize existing parts rather than inventing everything from scratch.
Living sponges are not unfinished humans. These studies do not establish one simple reason why sponges lack neurons.
The useful lesson is narrower: having components is not the same as organizing them into a new capability.
Humanity was thinking collectively before computers#
Language, writing, law, and scientific institutions already connect minds across distance and time. Knowledge can outlive its author. A hospital coordinates expertise that no single clinician could possess.
Collective intelligence did not arrive with a chatbot.
Digital networks changed the cost and speed of moving information. AI may change the cost of translating, comparing, and working with it.
Sending a report does not mean its readers understand it. Summarizing it does not make the summary accurate. Understanding the same risk does not produce agreement about sharing its costs.
A connected society therefore needs more than communication. It needs ways to preserve evidence, resolve disputes, distribute responsibility, and learn from consequences.
The internet shortened one kind of distance#
A large body needs coordination; a large society does too. But people are not neurons, and no global brain follows automatically from adding connections.
It helps to separate three distances.
Transmission distance concerns how quickly information travels. Understanding distance concerns whether people can interpret it. Interest distance concerns whether they want compatible outcomes.
The internet dramatically reduces the first. AI can sometimes help with the second. Neither automatically solves the third.
Every city could receive the same flood forecast and still disagree about releasing water from a reservoir. Uncertainty, unequal exposure to damage, and distrust would not disappear with more messages.
Simultaneous attention is not collective thought. A million people repeating a claim may be coordinating, copying, or making the same mistake. Speed cannot tell us which.
AI can help connect understanding, not manufacture agreement#
AI could help translate technical material, organize proposals, identify inconsistencies, and record decisions. These functions could make collaboration possible where its administrative burden previously made it impractical.
A bounded flood-response system might assemble observations, display forecast assumptions, and compare proposals. People could challenge its evidence before an authorized decision-maker acted.
But a forecast cannot determine a fair allocation of loss. That requires values, negotiation, and accountability. Predicting well does not confer the right to decide for everyone.
Adding AI does not always improve judgment, either.
A 2024 meta-analysis in Nature Human Behaviour examined 106 experimental studies. On average, human-AI combinations performed worse than whichever performed better alone: the human or the AI. They did improve on humans alone on average, a less demanding comparison.
The studies were published from January 2020 through June 2023. Results varied across tasks and designs. This is a warning against assuming automatic improvement, not a verdict on every current model or workflow.
The arrangement matters as much as the participants. Who checks sources? Who can reject recommendations? Which subtasks belong to whom? More intelligence needs a workable process to become better judgment.
A new species is a much stronger claim#
Biology describes transitions from separate units to higher-level organizations, including multicellular organisms. Connection alone is not the test.
A 2015 review of major evolutionary transitions discusses information transmission, cooperation, conflict, and new units of reproduction. The question is how a new organization forms and persists, not just how many components communicate.
The review already treats human societies with language as a major transition in a cultural sense. It also argues that human group structures are too transitory for the same conclusion in a purely biological sense.
Institutions transmit practices through education and imitation, but cultural transmission is not genetic inheritance. An increasingly coordinated society is not automatically a new reproductive individual.
Under the biological species concept, shared software and decisions do not establish reproductive isolation. They do not demonstrate that humanity has become a new species.
The AI-era question is whether technology could extend or reshape that cultural transition, not whether cooperation begins with machines. Could it strengthen collective agency: a durable ability to recognize shared problems, make commitments, act, and correct course across many people and institutions?
Shared memory is not enough without shared correction#
A collective system should preserve why it acted: evidence, uncertainty, objections, expected outcomes, and the conditions under which its decision should be reconsidered.
Otherwise, an archive can become a warehouse of decisions that nobody learns from.
Imagine a public project whose initial forecast is preserved alongside its results. AI could flag discrepancies. Learning only happens if somebody investigates them and changes the next decision. Generating a retrospective is not accepting its implications.
The engineering counterpart is designing responsibilities around AI-assisted work, not merely accelerating output. Societal design must also account for rights, competing interests, and people who did not choose to participate.
Evidence needs care too. The distinction between integrity and authenticity in file verification illustrates how one check cannot establish everything. An unchanged record can contain bad assumptions. A faithful summary can summarize weak evidence.
The test is not confidence. It is whether contrary evidence can reach people with power, change a decision, and improve the next one.
A larger “we” must not erase the individual#
Biological analogy is not political instruction. A person is not a disposable cell. How organisms constrain their components cannot justify eliminating dissent or demanding unlimited obedience.
A system that imposes one answer can become better at control without becoming better at judgment.
A better design would preserve independent evidence, appeals, privacy, and local decisions. Coordination should match the scale of a problem’s effects, not automatically favor the largest authority.
Differences can expose assumptions that uniform processes overlook. If everyone consults the same model, apparent agreement may reflect a shared blind spot rather than independent confirmation.
Building collective intelligence should therefore mean making cooperation possible despite disagreement, not making disagreement impossible.
Collective competence does not establish collective consciousness. None of the studies discussed here demonstrates that an internet-scale subjective experience exists or will emerge.
Make room for a civilization-sized pause#
The space between receiving a signal and acting can hold doubt and reconsideration. Collectively, it can hold evidence review, consultation, and a chance to notice an irreversible mistake.
This is a decision-making metaphor, not an account of how brains evolved. It is still a useful design goal.
Digital systems can compress that space until reaction outruns judgment. AI could accelerate the compression or help groups check claims, compare alternatives, and make uncertainty visible.
Emergency warnings need speed. Routine tasks need efficiency. Decisions with lasting public consequences need a process that fits their stakes.
Humanity’s next evolution need not mean everyone thinking alike. It could mean doing difficult things together without surrendering the right to question them.
Start smaller than a global brain. For your next consequential AI-assisted decision, record the evidence, uncertainty, accountable decision-maker, and condition that would make you reverse course. Return to that record when results arrive.
Before asking how quickly the system can act, ask whether it can still stop and say: what happens next?
Leave a Reply