I know many of you who've been reading my articles might think I'm against AI.
One cautionary tale after another, all I seem to talk about are risks, warnings, or how LLM is not ready for mass implementation.
Let me be clear. I'm bullish on AI, not the AI bullshit.
For 10 years, companies paid me to identify strategic intervention points to transform digital products. So it's only natural that whenever I research something, I see things we can improve or do better with AI.
AI, especially GenAI at this stage, is hardly a shovel for a gold rush. There is not yet a killer app that shows AI operating with full autonomy.
Today I want to discuss something I’ve observed.
This article has sat in my backlog for months now, I tweak it once every few weeks. Only now, I thought I had finally found a way to talk about this fascinating paradox.
When an AI produces something factually incorrect, we call it a hallucination, with a heavy tone of negative connotations. When AI produces something aligned with the request or even supersedes it, we call it imagination or creativity.
Underneath, these two are inseparable in GenAI.
And based on how it is currently built, you can't separate one from the other. Read this to understand the fundamental,
For millennia, we've paid premiums for human imagination. Novels, paintings, and scientific breakthroughs. Most of these products of the mind connected dots in ways that weren't strictly factual.
For example, I don’t think you will deny that the The Creation of Adam was a made-up scene?
There was never a live/dead cat in Schrodinger's lab, nor were Einstein's thought experiments initially grounded in experimental data. They were what-ifs that changed physics forever.
GenAI has kick-started the era of industrialized imagination. This doesn’t come without impact on how the world operated previously. This is a long story, and I will tell it in two parts.
So today, I want to bring a bit of history (because I know you enjoyed it), and why the industrialization of imagination isn’t necessarily a bad thing:
Hallucinations vs. Creativity. Human history is the thin line woven by both madmen and geniuses. And could Sam Altman be telling the truth, for once?
The similarities and differences between yours vs. GenAI’s Imaginations.
From life to pennies, stories of how imagination's cost has decreased due to technology.
Where is AI’s imagination champion?
Shall we?
Human Imaginations. Mad Men vs. Geniuses
Evolution selected hallucination as a survival strategy.
Your brain is a storyteller.
Every day, Klaas and I go for a walk after dinner. To reach the Thames' bank, we must cross a busy roundabout. When crossing roads, we humans don't constantly turn our heads 180 degrees to monitor traffic. Instead, our peripheral vision handles this naturally.
One evening, my visual cortex suddenly conjured an image of cars accelerating into my lane! My brain completed this phantom scenario with imagined speed and a collision trajectory, calculating whether the vehicle would strike me (not in the Sherlock Holmes way, of course).
Only, there was no actual swerving car.
It was just a plastic bag caught in the wind. Yet for that heart stopping moment, the phantom vehicle felt as real to me as any physical object. My body responded with tensed muscles and a spiked heart rate. All before my conscious mind could process what was actually happening.
This is pretty much how we perceive everything.
Neuroscientists at Sussex University have revealed that your visual cortex spends more time talking to itself than listening to your eyes.
As Professor Anil Seth explains,
Perception is never a direct window onto an objective reality, but rather active constructions, brain-based best guesses at the nature of a world that is forever obscured…
In plain English, your brain doesn't passively record reality. Your brain actively constructs it, filling gaps with its best predictions based on past experience.
Have you played with flip-book animations?
That's the phi phenomenon at work. When you saw two lights flashing in sequence, you saw one light moving between positions. Rather than two separate events. Your brain was inventing motion that never existed. The same reason movies work at all.
You and I hallucinate many things we thought we’d observed.
It was exactly these false positives that kept our ancestors alive.
Those who imagined threats in ambiguous shadows survived to become our ancestors; those who demanded evidence became lunch.
The anxiety you feel walking alone at night, when you're sure someone's following you?
It's a feature, not a bug.
The Shortcut to Imagination: Getting High
Many of us aren’t satisfied with just our natural imaginative capacities.
From ancient shamans chewing ayahuasca in the Amazon, Francis Crick envisioned the double-helix structure of DNA while on LSD, Steve Jobs dropping acid in Silicon Valley… no doubt, a lot of us will continue to induce hallucinations to break mental patterns deliberately.
Conscious or not, these people get that imagination often requires breaking from consensus reality. We've built rituals around controlled hallucination as a pathway to insight.
The "Ridiculous" Ideas That Changed Everything
The Earth moves around the Sun—Copernicus (considered absurd)
or the
Invisible 'germs' cause disease—Semmelweis (mocked and institutionalized)
What about…
Time slows down when you move faster—Einstein (initially rejected by physics establishment)
Were these moments of hallucination or imagination? The distinction isn't in the mental process, but the utility of the output.
Maybe the most valuable outputs aren't always the factual ones but the wild, unexpected connections no human would have ever made?
After all, most of our progress came from people who saw things that weren't there yet.
The Striking Similarity, Yours vs. GenAI’s Imaginations.
What are the shared malfunctions between our brains and AI systems?
The same predictive processing failure that causes both GPT-4 to fabricate citations and your brain to see faces in clouds.
Ever made up an explanation when you can't remember something exactly? Both you and AI face data sparsity challenges, confabulating to fill gaps.
That time you "remembered" a childhood event that never happened? You are like AI, which suffers from reward-driven overfitting. Prioritizing coherent narratives over accuracy, and psychological comfort rather than truth.
When you hyperfocus on a conspiracy theory's supporting details (eg, flat Earth) while ignoring contradictory evidence, you're experiencing the same attention mechanism dysregulation that causes AI to fixate on certain patterns while ignoring others.
Again, instead of storing complete memories, your brain reconstructs them from fragments. That's why eyewitness testimony is notoriously unreliable.
The Differences, Yours vs. AI Imaginations
While I've covered the similarities between human and AI hallucination, the differences are where the real economic and cultural implications emerge. These fault lines will determine which forms of imagination retain particular value in an age of abundance.
The Intentionality
Human imagination is born from intention.
When I write to you, I'm driven by the need to express, to be understood, to connect.
It was certainly beyond pattern completion. I imagine because I want something to exist that doesn't yet.
I watched my goddaughter construct an elaborate fantasy world with a marker. "This is a rocket ship," she said, "but it runs on rainbow fuel, and only unicorn farts can make rainbows."
That was not a random pattern generation. It was imagination with purpose, reflecting her desires to share the world she saw.
GenAI imagination lacks this intentionality. As I mentioned in this article:
GenAI doesn't create because it wants the creation to exist. It generates outputs based on statistical patterns without the desires that drive human creativity.
The Nature And The Origins of Data (Experience)
Human imagination thrives within constraints. Same as GenAI.
Your creativity is never limitless; no one is.
Your experiences, cultural background, physical body, and cognitive limitations shape your imagination. This background also restricts your imagination; it gives it shape and meaning.
This nature of human creativity means meaningful boundaries are shaped by lived experience. So, even between you and I, we’d have very different imagination constraints.
GenAI also has constraints.
But they differ fundamentally in both origin and nature.
While AI's limitations stem from its training data, which is essentially an aggregate of human-created content. AI’s boundary is the human collective knowledge.
It sounds large and abundant, yes.
However, imagine there are 10 different people inside you, and you have all the memories, you’d be confused about the origins very soon.
That’s why you often see AI have trouble referencing, to distinguish between factual and fictional. AI cites nonexistent studies, quotes fake statistics, and creates plausible-sounding sources that never existed.
The Connection With the Physical World
My imagination lives in my body.
When I imagine running, my heart rate increases slightly. When I imagine tasting something sour, my salivary glands respond. My imagination is inseparable from my physical existence.
Last year, I spent nights in a Texas Hold'em tournament. When I’ve sat for hours calculating pot odds while my heart races, watching opponents for the slightest twitch that betrays their hand, I develop this relationship with uncertainty that no algorithm could understand.
Each decision carries the weight of real consequences. It was more than money, but imagining the victory, the thrill of seeing the full room of men defeated. Well, only if you’re curious, my imagined strategy didn’t work out.
GenAI’s imagination floats in a realm of pure pattern; there’s no anchor.
It has never felt hunger or exhaustion or desire. It has never experienced the world through senses. This disembodiment creates a fundamental limitation. As GenAI can mimic the outputs of embodied imagination, but cannot replicate the process that gives those outputs their depth and resonance.
As Professor Robert Root-Bernstein explains in the Journal of Creativity,
Most human creativity is embodied and involves the manipulation of tools and materials… current AI systems employ a very limited range of mental skills, especially those associated with imagination and intuition that rely on perceptual, sensory and emotional feelings or requiring material interaction with nature or the environment.
Point Out Facts From Fictions
Perhaps the most significant difference lies in metacognition.
Our ability to think about our own thinking.
I know when I'm being creative versus when I'm recalling. I can tell you which parts of my work feel derivative and which feel original. I can assess my confidence in different aspects of my creative output.
When writing this article, I'm constantly evaluating my own ideas. If a metaphor seems forced, I'll find a better one, or when an argument needs stronger evidence, I'll research further. This metacognitive process is part of my creativity process.
Current AI lacks this metacognitive layer.
It cannot distinguish between what it knows and what it's fabricating. It cannot tell you which parts of its output are derivative and which are novel combinations. It cannot assess the quality of its own imagination.
From Imagination Scarceity To Surplus
Imagination is the spark that starts everything to follow suit.
When only Portuguese sailors could imagine routes around Africa, they tried and found the route, then they controlled Asian trade; Bell's telephone empire started from imagining long-distance voice transmission; or Jobs imagined a personal computer with a friendly interface…
You can't create what you can't first imagine.
You can't innovate toward a future you can't envision. You’d have no idea of solutions to problems without first imagining alternatives.
So…
What happens when the most valuable thing GenAI produces is imagination?
I rarely find myself agreeing with Scam Altman (oops, typo).
But his perspective on hallucination might be one of his rare, non-salesy takes.
… the fact that these AI systems can come up with new ideas and be creative, that’s a lot of the power… if you just do the naive thing and say ‘never say anything that you’re not 100% sure about’, you can get them all to do that. But it won’t have the magic that people like so much. — Salesforces interview 2023
The same mechanisms that allow AI to hallucinate are precisely what enable it to imagine new drug compounds, design novel materials, or generate creative solutions.
There’s no creativity without hallucination from AI. They're inseparable functions, two sides of the same computational coin.
Many are obsessed with the factual-fictional or the human vs. AI divide.
However, these do not matter to the market. The majority of consumers don’t care about any of those. Utility is the key.
Imagination has been humanity's most jealously guarded treasure. Until now:
The same mechanism that produces "hallucinations" also produces valuable innovations
GenAI is industrializing imagination at scale, ending imagination scarcity
This collapses the imagination premium, which was previously rare and costly to produce
Some stories (history) might help you understand the transition.
The Cost of Imaging New Materials
In 1500, in Germany, an alchemist worked for a duke who imagined and promised that he knew how to turn iron into gold. Two tons of iron were hauled in at great expense.
Soon, after a few failed attempts, the duke's patience ran out.
Sensing danger, the alchemist fled Germany. But not far enough.
After his capture, the duke ordered a spectacle like no other.
The alchemist was dressed in a garment dripping with gold tinsel. On a hillside visible to all, a special gallows awaited. The duke had ordered it built from the amount of iron the alchemist had promised to transform into precious metal.
The duke made sure everyone knew what happens when a promise is made with impossible imagination and fails to deliver.
The entire structure was coated in gold leaf, gleaming in the sun as a final insult.
In medieval times, the cost of a failed imagination was your life.
This was not a standalone case.
Another alchemist lost his head in Munich after promising a duke the secret of gold-making. It was said that the executioner had to use three strikes with the sword until his head was eventually cut off.
Material imagination remained scarce during the Industrial Revolution. Each breakthrough still took decades of trial and error.
For example, Sir. Bessemer imagined and realized a brand-new method to remove unwanted materials in steel production.
Before this innovation, converting iron to steel would take at least a full day of heating, stirring, and reheating. In contrast, the Bessemer process reduced this to 20 minutes to convert tons of iron into steel. The records showed the process decreased costs from £40 per ton to £6–7 per ton.
Today.
DeepMind GNoME is an AI system that has discovered 2.2 million new crystal structures in a single project (months as a unit). Not modifications. Not variations. But brand new structures. Among these, 380,000 are stable enough for practical use, brand new materials that no human had imagined.
This graph from DeepMind should tell you how impressive this achievement is.
That's equivalent to 800 years of human discovery, delivered in months.
With an imagination surplus, the material engineers around the world couldn’t catch up to verify and synthesize all these potential beneficial materials. Labs worldwide have synthesized 736 out of 380,000 AI-imagined materials, confirming what the machine dreamed of.
A fascinating example of how AI can actually help improve our environment or quality of life.
The Cost For You To Dress Like A Royal
A single dress once cost more than a peasant earned in ten years.
In Tudor England, one Queen Elizabeth’s gown alone used fabric worth £25,000 in today's money, and over 1000 labor hours to craft. Neither of us was likely to afford this if we were born then.
Or when Charles Worth, a famous fashion designer (think of Karl Lagerfeld, creative director of Chanel from the 1980s until his death), opened his Paris salon in the 1850s, empresses waited months for his attention. 70% of his clients are royalty or aristocracy.
Again, if you weren’t born that rank, you likely couldn’t afford it.
His ability to imagine new silhouettes made him fashion's first dictator. When one man's imagination clothed the world's elite.
Since 2023, AI-designed garments cost 90% less to prototype than human counterparts. Equally, the global creative AI market grew to $62.5 billion in 2024, even as freelance designer wages fell 22% in the same report.
Today, you have DALL-E and Midjourney produce more fashion concepts in an hour than Worth created in his lifetime.
What once required royal patronage now requires only a prompt (click).
The most jealously guarded imagination in history has become commonplace.
The market value of raw imagination is free-falling wherever AI operates because of the marginal utility; unless you care for the logo and brand names, you can buy an extremely similar style on Shein or Temu as opposed to paying thousands of dollars to Chanel or Dior.
In addition to AI-designed clothes, the cost of shopping has also decreased for merchants and customers. At Google I/O, they released the try-on feature that lets anyone instantly see themselves in clothes.
Unlike material discovery, the pros and cons of this change are much harder to determine.
All we know is that it is harder for a designer to stand out, and the imagination itself is commoditized.
Where AI Imagination Thrives
The economics of AI imagination follows a clear pattern across industries.
Here are two contrasting themes in front of you, facing the same technological revolution:
Similar to material discovery, pharmaceutical research, GenAI presents potential drug candidates in seconds. However, that's merely the first step in a two-decade conversation with reality. Or using AI to help review X-ray and MRI results.
Similar to cloth design, in digital marketing, AI generates campaign concepts, copy variations, and visual designs. Then the result can be deployed, tested, and optimized within hours.
Both utilized AI’s imagination. Wildly different impacts.
Why is that?
To explain this, let’s go back to the difference between AI and us.
Pattern Recognition vs. Embodied Experience
AI imagination excels at pattern recognition across vast datasets. It can identify and recombine elements from millions of examples, generating variations at a scale no human could match.
This gives AI decisive advantages in domains where:
Patterns are clear and well-documented
Success metrics are straightforward to quantify
Emotional resonance is secondary to functional utility
This explains why AI rapidly transforms industries like digital marketing, where success metrics are clear (engaging posts had high conversion rates and click-through percentages), patterns are abundant in historical campaign data, and there are limited emotional connections with those who view the ads (I’m not talking about the NFL ads, but the cheap social media advertising alike).
Your imagination, again (I will discuss in part 2), by contrast, draws on embodied experience. The lived sensations, emotions, and cultural contexts that shape our thinking. All good things to have, but costly.
Iteration Speed vs. Metacognitive Judgment
AI imagination operates at computational speed, generating and testing thousands of variations in seconds. This creates overwhelming advantages in industries where:
Rapid iteration improves outcomes
Verification is immediate and low-cost
Volume of options increases the success probability
This explains AI's rapid adoption in content creation, where producing multiple variations costs virtually nothing and immediate feedback loops enable quick optimization.
Even with metacognitive awareness, human imagination comes much slower. The fact that we know what we know and what we don't isn’t necessarily an advantage under this category.
Data Dependency vs. Novel Insight
AI imagination is fundamentally data-dependent, excelling where abundant examples exist but struggling with true novelty. This creates clear industry dividing lines:
Industries with data abundance see rapid AI adoption. Especially because the current models are trained on publicly accessible data. The more examples available, the more effectively AI can generate new variations. That’s why you see content creation, visual design, and routine coding are the tasks that benefit from the vast data access.
And AI struggles when you ask domain-specific questions, the deeper you go, the more mistakes you’ll find.
There’s only so much you can do by mixing beans and toast. The same principle applies when it comes to breakthroughs and innovation. So that initial spark of imagination allows absolutely new concepts that might still require humans.
I said might… because… for example, DeepMind's protein folding breakthrough came from patterns humans never saw. Even these are still recombining existing knowledge, not the leap-of-faith insights that create entirely new fields of study.
The border is getting blurred.
The Human to AI Imagination Transformation Threshold
These differences create a clear transformation threshold. Industries and tasks transform rapidly when they feature:
Low verification costs for acceptable GenAI outputs
Abundant data that provides rich patterns for AI to learn from
Clear success metrics that allow for objective evaluation
Low stakes for failures during iteration
Limited need for emotional or physical resonance
When these conditions exist, the transformation is not gradual but sudden and comprehensive. As they do in digital marketing, content creation, language learning, and basic consultation (see my AI Wins 64% of Debates Against You! for context) etc.
When these conditions are absent, human imagination retains its premium value. Just as we’ve discussed, in scientific research, high-stakes medical decisions, and medicinal examinations.
The Age of Industrialized Imagination
For the first time in human history, imagination, which is supposed to be a unique talent to humans, is no longer ours alone.
It has been industrialized and commoditized… to an extent.
Some part of the imagination is now reproducible at near-zero marginal cost. AI can scale without precedent, can recognize patterns across impossible domains, is tireless, and can iterate limitlessly.
When verification costs are low, data is abundant, metrics are clear, stakes for failure are manageable, and emotional resonance is secondary… AI imagination possesses many advantages over us.
The economics are undeniable. We’ve observed this in content creation, visual design, software development (ish), AI audio cloning, not to mention document processing tasks like summarizing text, translation, and so on.
When imagination becomes infinitely reproducible, industries built on the scarcity of creative talent must restructure or face obsolescence.
Before we end this one today, I have some questions for you.
Now, imagination becomes abundant through AI. Does judgment become the new scarcity? (The ability to determine which imagined possibilities are worth pursuing)
Does contextual imagination, understanding which solutions fit specific cultural, ethical, or human contexts, become the premium skill?
Finally, who profits when the price of imagination approaches zero?
Share this post