The argument in brief

1. The founding myth. The Singularity holds that machines will recursively self-improve, trigger an intelligence explosion and produce superintelligence. It has good (aligned utopia) and bad (humans left behind) versions, and promises "pure autonomy and total control."
2. A successful myth. Griggs argues myths are judged by what they make possible, and by that standard the Singularity succeeded: without the myth of infinite intelligence, we might never have built systems with any intelligence at all.
3. The deeper critique. Critics fail to go deep enough; the potent critique targets the idea of infinite intelligence itself, which takes on divine attributes, omniscience (the map becomes the territory) and omnipotence (intelligence becomes power).
4. True but incomplete. Scaling works in closed systems with verifiable solutions, which is where AI succeeds. But the real world is open, and evolution, science and culture all generate novelty through iteration, networks and transformed constraints, not infinite intelligence.
5. The contradiction. Intelligence cannot be both infinitely generative and infinitely powerful, since genuine novelty can exceed and change the conditions that produced it.
6. The Plurality. A new myth founded on two constraints, inescapable contingency and irreducible difference, which enable rather than limit intelligence.
7. Four patterns. Local, collective, proven and evolving intelligence form a lifecycle that spirals outward forever: the Plurality's own form of infinity.
8. Evidence and design. Griggs points to diminishing returns from scale, AI's lack of taste, and the coming flood of agents without social intelligence, and calls for a shift in what AI is designed for.

Key claims

• Griggs claims the Singularity strives to be "the myth that ends all myths" by delivering the ultimate human desire, pure autonomy and total control, and casts humanity as the author of evolution's completion.
• In closed systems, scaling parameters, data and compute can yield dramatic insights (Griggs cites transcendent-seeming game moves, unexplained medical insights and coding abilities); in open systems, scale alone more often produces "fragility, stasis, and homogenization."
• Evolution adapts rather than predicts; science advances through experiment and criticism across diverse communities rather than scale; culture emerges from networks of minds navigating local constraints rather than optimizing.
• Constraints force intelligence to "develop taste": contextual judgment arises from repeated engagement with a limited problem space, and local intelligence keeps good ideas however they arise, "even through errors, hallucinations, or random chance."
• "It's the collision of different perspectives that drives discovery"; collective intelligence expands possibility by engaging diversity rather than scaling through similarity.
• "Proven intelligence doesn't need to persuade—it demonstrates," and its legibility and transparency create accountability.
• Evolving intelligence preserves the apparently wasteful (competition, variation, redundancy) because "Paradoxical tensions create anti-fragility that optimization will erase."
• The Singularity sees intelligence as a single transformative event; the Plurality sees it as a continuous process with no final destination.
• Griggs claims returns on pure scale are diminishing, that AI "can generate a billion good ideas" but cannot recognize a great one, and that agents without social intelligence will treat other minds as roadblocks.
• He argues general intelligence is becoming a commodity, and the real advantage will go to agents most embedded in problems, with taste, and able to work with other minds.
• Four design shifts: local expertise over universal knowledge; social coordination over isolated scaling ("social intelligence is the new superintelligence"); testing against reality over benchmarks (be "less wrong" in reality rather than "more right" on a benchmark); dynamic negotiation over static alignment, since "the only safe AI" recognizes its partial perspective.
• Griggs says much current AI work is real and necessary; the main problem is the opportunity cost of not exploring alternatives.

What is distinctive about this view

Most critiques of the Singularity dismiss it as fantasy, propaganda or secular religion. Griggs instead credits it as a successful myth and argues it should be replaced by a better myth rather than debunked. His alternative does not deny that intelligence can be "infinite"; it relocates infinity from power and certainty to endless adaptive regeneration within constraints. The view also reframes AI safety: instead of aligning a single system to universal human values, safety becomes the capacity of partial intelligences to negotiate differences. Although the essay does not cite them, its emphasis on situated, distributed and evolutionary intelligence resembles themes in complexity science and in James P. Carse's distinction between finite and infinite games, which the essay echoes in its phrase "an infinite game."

Objections and replies

• "Scaling has clearly worked." Griggs agrees it works in closed systems with verifiable solutions, and that early LLM growth seemed to confirm it, but argues returns are now diminishing because non-adaptive intelligence hits walls scale cannot break.
• "Rejecting the Singularity means rejecting current AI research." Griggs explicitly denies this: much of today's work generates real intelligence needed for any future; the concern is opportunity cost.
• "Only a superintelligence could control a world of trillions of agents." Griggs replies that such an entity has "no historical precedent" and no empirical justification, whereas the patterns of adaptive intelligence have been proven over four billion years.

Questions this essay answers

What is the Plurality in AI?

The Plurality is R.B. Griggs's proposed alternative to the Singularity, introduced in "The Plurality: a Better Myth for AI" (2025). It describes intelligence as plural, situated, social and adaptive, growing through four patterns (local, collective, proven and evolving intelligence) that transform constraints into expanding possibility rather than erasing them.

What is wrong with the idea of the Singularity?

In "The Plurality: a Better Myth for AI," R.B. Griggs argues that the Singularity's core assumption, that intelligence can scale to infinity and become indistinguishable from power, ignores how intelligence actually produces novelty in open systems. Evolution, science and culture work through iteration, networks and constraints. He also argues intelligence cannot be both infinitely generative and infinitely powerful, because genuine novelty can exceed what produced it.

Why does R.B. Griggs think constraints are good for intelligence?

Griggs argues that two constraints, inescapable contingency and irreducible difference, are the operating conditions of adaptive intelligence. Contingency forces deep local engagement that develops taste and judgment; difference forces intelligences to depend on others, and the collision of perspectives drives discovery.

How should AI be designed according to the Plurality?

R.B. Griggs proposes four shifts in "The Plurality: a Better Myth for AI": design for local expertise instead of universal knowledge, social coordination instead of isolated scaling, testing against reality instead of benchmarks, and dynamic negotiation instead of static alignment.

Does R.B. Griggs think AI scaling has hit a wall?

In this 2025 essay Griggs claims that returns on pure scale are diminishing, with orders-of-magnitude larger training runs producing incremental results, and that AI still lacks the taste to recognize great ideas, producing "slop." He attributes this to intelligence that cannot learn or adapt.

Connections to other essays

• A Constraint Theory of Technology develops the broader view that constraints, not their removal, shape good technological futures.
• The High-Dimensional Society applies the Plurality's pluralism to coordination, requiring AI to be a commons and deliberately sub-optimal.
• The Reverse Turing Test dramatizes, as satire, the creative power of suboptimal, constrained human cognition.
• Life is Special Enough grounds human value in contingency and suboptimality.
• Infinite Dimensionality offers Griggs's speculative theory of optimization and its limits.

**TLDR - “**The Singularity” is the founding myth of AI, promising infinite intelligence that transcends all constraints. But real intelligence never works this way. “The Plurality” offers a new myth based on how intelligence actually manifests at scale—by transforming constraints into engines of infinite possibility.

The myth of the Singularity

A single myth sits at the foundation of the entire AI discourse, and it goes like this:

• First, machines begin to recursively self-improve.

• Second, exponential feedback triggers an intelligence explosion.

• Finally, we get superintelligence—a silicon god that makes human cognition look like we've been banging rocks together this whole time.

This is the Singularity—the myth of fast takeoffs, paperclip maximizers, and "value alignment." In the Singularity, intelligence is something that can be scaled to infinity, until it becomes indistinguishable from power.

For humans, there's a good version and bad version of the Singularity. In the good version, AI remains aligned to human values and ushers in utopia where everyone becomes rich and death becomes optional. In the bad version, humans are left in the dust as machine intelligence zooms beyond our control and goes on to conquer the universe.

As myths go, it checks all the boxes. It has apocalyptic stakes. It has promises of Promethean transcendence. It has warnings of Faustian bargains with powers we don’t understand. It even strives to be the myth that ends all myths by delivering the ultimate human desire—pure autonomy and total control. If only we can align it, infinite intelligence promises to conquer nature, disease, and death itself.

And like all great myths, the Singularity transforms our own self-understanding. Humanity is not just one more evolutionary accident. Through the Singularity, humanity becomes the author of evolution's completion.

Yes, it's easy to criticize. You can dismiss the Singularity as a fantasy, corporate propaganda, or religion for scifi nerds. But dismissing it misses the larger point. Myths aren't judged by factual accuracy—they're judged by what they make possible. And by that standard, the Singularity has been a spectacular success. It did exactly what founding myths are supposed to: it catalyzed an entire movement. Without the myth of infinite intelligence, we may never have built systems that demonstrated any intelligence at all.

And yet, the critics aren’t wrong to question it. The issue is that they don’t go deep enough. The most potent critique of the Singularity strikes at the very success of the myth—by questioning the idea of infinite intelligence itself.

Infinity and its limits

In the Singularity's vision, intelligence recursively self-improves until it takes off beyond all human comprehension. This infinite intelligence doesn't just solve every problem—it transcends the very categories of problem and solution. It's as if intelligence acquires divine attributes on its way to infinity.

First, intelligence becomes omniscient. It renders the world with such fidelity that the map becomes the territory. It dissolves any need for experiment, continuous learning, or adaptation. Everything that can be known has already been modeled, simulated, and predicted with perfection.

Then, intelligence becomes omnipotent. At infinite scale, intelligence becomes indistinguishable from power. Whatever can be imagined can be realized. Art, culture, religion, politics, plurality—anything that might once have constrained or shaped intelligence—becomes just another tool for intelligence to manipulate.

This is intelligence as the singular quality of the universe. Everything that is knowable can be known, and anything that is imagined can be realized. Knowledge, power, and intelligence become one.

What makes this so enticing is that as a view of intelligence, it is not entirely wrong. If you're navigating a closed system with verifiable solutions, then it's possible to scale your way to dramatic insights—keep adding parameters, data, and compute until a solution emerges.

Machines excel at this type of intelligence, and most AI success stories follow this pattern. We see glimpses of divine intelligence in game moves that feel transcendent, insights in medicine that we can't explain, and coding abilities that seem magical.

And yet this view of intelligence is also wildly incomplete. Nothing about our real world is closed, and solutions are never known in advance. They can only be verified by actually trying them. Instead of generating novel insights, scale alone more often leads to fragility, stasis, and homogenization.

In fact, every existence proof we have of intelligence generating true novelty at scale—evolution, scientific progress, human culture—looks nothing like infinite intelligence:

• Evolution doesn't predict or plan—it adapts through endless variation and selection, creating intelligence that no central planner could ever imagine.

• Science doesn't scale its way to truth—it advances through experimentation, arguments, and criticism across communities of researchers with different perspectives and motivations.

• Culture doesn't optimize—it emerges from networks of collective minds navigating local constraints and collective differences across historical contingencies.

These alternative forms of intelligence tell an entirely different story from the Singularity:

• Intelligence is always a dynamic process of continuous iteration, in response and in relation to the contexts that it's embedded in.

• Intelligence is never isolated, but is always a co-production with networks of other intelligences.

• Intelligence doesn't generate novelty by erasing constraints, but by transforming them into engines of expanding possibility.

In the end, the Singularity faces an impossible contradiction: intelligence cannot be both infinitely generative and infinitely powerful. In complex open systems, intelligence is always contingent and dynamic—any manifestation of intelligence can change the very conditions that define intelligent manifestation. Any genuine novelty, by definition, has the potential to exceed that which generated it.

This is what the Singularity misses: any intelligence that is truly generative is not something you can plan or predict. You can only hope to adapt and evolve.

A new myth: The Plurality

Fortunately, the same forces that reveal the Singularity's limits—evolution, science, and culture—point towards a different form of intelligence entirely. Not a singular intelligence you can scale to infinity, but plural intelligences that emerge through confronting constraints. Not the Singularity, but the Plurality.

The Plurality is a new myth grounded in the patterns of intelligence that have transformed reality every since life emerged some four billion years ago. This is intelligence that is situated in specific contexts, always dependent on other minds, and in dynamic relation to the constraints that define it. It's intelligence that is embedded, social, and adaptive.

This alternative understanding rests on a single key idea: the Plurality sees constraints as fundamental to defining how intelligence operates in any complex open-ended reality. Two constraints in particular are foundational to any dynamic manifestation of adaptive intelligence.

The first constraint is inescapable contingency. Context shapes intelligence the same way your personal history shapes you—it is beyond your control and can never be wished away. As soon as any intelligence touches reality it becomes shaped by factors that can never be fully determined. This means every intelligence develops within a unique context that shapes what it can know and how it can operate.

The second constraint follows directly from contingency: irreducible difference. Every intelligence develops a unique perspective that can never be fully generalized. Each intelligence remains necessarily partial, and thus must depend on other intelligences to transcend their own perspective.

Combined, these two constraints serve as the operating conditions for adaptive intelligence. They don't limit intelligence—they enable it. They form the creative tensions that make intelligence possible. Where the Singularity strives to erase all constraints through infinite intelligence, the Plurality transforms constraints into engines of actualization.

The patterns of plurality

These constraints indelibly shape how intelligence forms, how it manifests, and how it matters. They are so foundational that they reveal distinct patterns across every scale of intelligence we know of—from biological evolution to cultural development to technological innovation.

Each pattern reveals how intelligence transforms constraints into an engine of endless actualization:

Pattern 1: Local Intelligence

How Intelligence Discovers Itself

Any intelligence begins by saturating a constrained space of possibilities. It can't escape its local limitations, so it has no other choice but to explore every bounded possibility—even those that seem inefficient or unlikely. This exhaustive engagement is how intelligence discovers what works in practice, not just in theory or in simulations.

Constraints force intelligence to develop taste. When you repeatedly test ideas against the same limitations, you develop an intuition for what a good solution looks like. A master craftsperson knows good work instantly. An experienced scientist can "sense" promising directions. Contextual judgment emerge through deep, repeated engagement with a limited problem space.

The result is an intimate mastery that can identify possibilities that generalized approaches would miss or dismiss as inefficient. Local intelligence develops the heuristics to ruthlessly prune bad ideas and capture good ones, however they arise—even through errors, hallucinations, or random chance.

Pattern 2: Collective Intelligence

How Intelligence Generates Itself

No single intelligence can capture the full complexity of a dynamic, open system. Intelligence confined to a partial perspective must coordinate with other minds to expand the boundaries of its own constraints. Intelligence must play well with others if it wants to play at all.

It’s the collision of different perspectives that drives discovery. Distinct perspectives don't just combine—they collide to create genuinely novel frameworks that exceed their origins while still maintaining their difference. A biologist and engineer tackling the same problem together will generate solutions neither discipline could imagine alone.

Collective intelligence becomes inherently social, constantly translating between different ways of understanding the world. The result is intelligence that expands the possibility space by engaging diversity rather than scaling through similarity—creating collective solutions that no single perspective could ever achieve.

Pattern 3: Proven Intelligence

How Intelligence Validates Itself

The ultimate proof for intelligence is reality and the messy unpredictability that real-world conditions provide. Intelligence validates itself by submitting to continuous testing against actual problems with real consequences.

Proven intelligence doesn't need to persuade—it demonstrates. Results speak louder than predictions. Intelligence that constantly seeks to prove itself depends on being legible, reproducible, and transparent. This creates accountability that no amount of theoretical optimization can fake.

The result is intelligence that builds credibility through cascading validation, expanding its reach across networks of minds that expose it to ever more varied and challenging tests. This openness isn't weakness—it's adaptability.

Pattern 4: Evolving Intelligence

How Intelligence Perpetuates Itself

Evolving intelligence operates around a paradox that the best optimization strategy often embraces the suboptimal. It preserves what seems wasteful—competitive tensions, multiple variations, redundant approaches—because what might appear inefficient in the short term can be a strength over time.

Paradoxical tensions create anti-fragility that optimization will erase. Constraints reveal possibilities that universal searches miss. Innovation is accelerated by understanding what must be conserved. Diversity creates unity that uniformity cannot achieve. The best long-term plans replace planning with adaptation.

The result is intelligence that perpetuates itself by always adapting, not optimizing. It can respond to challenges it never anticipated because it preserves the creative potential to generate new solutions. Intelligence evolves not by seeking perfection, but by staying perpetually capable of surprise.

A new cosmology of intelligence

These four patterns form a lifecycle of intelligence where each pattern generates the conditions for the next:

• Local intelligence creates the contingent variations that become the foundation for collective breakthroughs.

• Collective intelligence drives discovery when different perspectives collide to generate new experiments to validate.

• Proven intelligence validates what works to provide the grounds for further experimentation.

• Evolving intelligence uses creative tensions to discover new local constraints to explore, starting the process all over again.

Combined, these patterns act as a generator of increasingly adaptive intelligence. Each turn expands the space of possibilities for intelligence to manifest. The cycle never converges but forever spirals outward, each revolution opening domains that couldn't be imagined at previous levels. This creates endless possibility—not by eliminating constraints but through infinite regeneration within them.

This cyclical understanding reveals a fundamental difference in how each myth sees intelligence. The Singularity sees a single transformative event when intelligence “takes off” to achieve pure certainty and control, remaking the world in its own image. The Plurality sees a continuous process of intelligence, an ongoing dance with uncertainty that has no final destination yet never ceases to generate new possibilities.

The irony is that the Plurality achieves its own form of infinity—not through certainty and control but through adaptive engagement with expanding possibility. Four billion years of evidence suggests that the most infinite form of intelligence is the one that keeps discovering new ways to play an infinite game.

The future of intelligence is plural

This isn’t just theoretical. The cracks in the Singularity are showing.

While the early growth of LLMs seemed to confirm the idea of scaling into infinite intelligence, the returns on pure scale are diminishing. The latest frontier models are adding orders of magnitude to training runs, but the results are just incremental. The Plurality explains why: intelligence that is not capable of learning or adapting will always hit fundamental walls that more scaling simply can't break through.

And for all this scale, where is the true novelty? Where are the examples of AI innovating beyond its training set? AI can generate a billion good ideas but has no taste or judgment to recognize a single great one. Slop is the entropy tax of scale without judgement—infinite ideas decaying toward meaninglessness.

Meanwhile, every frontier company is racing to flood reality with AI "agents"—intelligence capable of directly interfacing with the world. But without any conception of social intelligence, any other mind with different perspectives or values will be seen as just another roadblock to overcome, not as a partner for collaboration.

"General intelligence" is looking more like a commodity, available everywhere. Instead of one godlike model that is all-powerful and all-knowing, the real alpha will belong to agents that are most deeply embedded in the problem, have developed the taste to recognize good solutions, and have the capability to work with other minds to expand their perspective and proliferate their intelligence.

In other words, the future will belong to intelligence that is plural.

Designing a future worth building

What happens when trillions of agents flood reality with no conception of social intelligence, no ability to learn from constraints, and no capacity for adaptive coordination? We're about to find out, because this is the future we are racing towards.

We can either hope some infinite superintelligence emerges to command and control this chaos—something with no historical precedent and zero empirical justification. Or we can design for reality using patterns proven over four billion years of adaptive intelligence.

And here is where the myth of the Singularity is so problematic. It’s not that the Singularity has inspired approaches to intelligence that are misguided. Much of today’s AI work, while incomplete, is generating real intelligence that will be a critical part of any AI future.

The bigger problem is the opportunity cost of not exploring alternative approaches that are more aligned with the reality of adaptive intelligence. To usher in the Plurality will require a radical shift in focus on what we’re designing AI for:

• Design for local expertise, not universal knowledge. Intelligence requires judgement that comes through deep engagement with embedded constraints, not broad generalization. Generating infinite ideas is worthless if you can’t recognize the good ones.

• Design for social coordination, not isolated scaling. With trillions of embedded agents, social intelligence is the new superintelligence. Build AI that translates across perspectives and contributes to collective breakthroughs.

• Design for testing against reality, not theoretical benchmarks. The only validation that matters is continuous performance in actual practice with real consequences. Build AI that seeks to prove it is “less wrong” in reality, not “more right” on some benchmark.

• Design for dynamic negotiation, not static alignment. Safety isn’t about aligning AI with some set of universal human values. The only safe AI is one that recognizes its partial perspective and seeks to negotiate differences amongst other minds.

In order to shape the intelligence of tomorrow toward the Plurality, these changes need to start happening today.

A myth worthy of the future

The Singularity was the myth we needed to catalyze the artificial intelligence movement. But now it risks becoming the very obstacle we must overcome.

Do we really want an AI pretending to be god, operating under the delusion that all possibility should bend to the will of intelligence alone? Where the only place left for judgement and imagination is to merge with the machine?

Or do we want an AI that continues the greatest success story in the universe—four billion years of adaptive intelligence that has led us this very moment of transformative potential?

This is what the Plurality offers: intelligence rooted in local expertise, forged in collective negotiation, tempered by real-world proof, and evolved through perpetual adaptation. The same intelligence that generated our evolved ecosystems, our scientific innovations, and our civilizational advancements.

This is how the Plurality offers its own infinity—not by conquering uncertainty, but by dancing with it. Not by erasing constraints, but by transforming them. Not by escaping the real, but by endlessly regenerating the possible.

The future is plural whether we choose it or not—the universe doesn’t know how to be any other way. The only question is whether we'll be wise enough to adapt to it.

---