From the Wheel to the Algorithm: Why We Panic Over New Tools
In this episode of The Sanity Project, we’re diving into the roots of outrage culture, exploring how a historical lens can deepen our critical thinking amid today’s ceaseless flood of news and information. Our news breakdown unpacks the AI “slop” phenomenon—showing that, while the technology is new, our panics and reactions are nothing of the sort. Join us as we question received wisdom and look at current events from a perspective grounded in centuries of change.
To subscribe to our free weekly newsletter, go to https://thesanity.org/subscribe
Understanding Outrage Culture in Canadian and Global Politics A Critical Lens on Liberal DemocracyAmid the rapid churn of daily news, it’s easy for outrage culture and media misinformation to drown out reasoned political analysis. Progressive politics and liberal perspectives call for robust critical thinking—challenging reactionary narratives while advocating for transparency and accuracy in news commentary. In a democratic society like Canada, this means holding our institutions and media to account.
Canadian News, Politics, and Media AccountabilityFrom Canadian news channels to Canadian politics, questions about democratic values, misinformation, and news manipulation often dominate the headlines. Our political commentary pushes past the surface-level noise of the latest current events—offering clear-eyed news analysis grounded in data, history, and progressive ideals.
Navigating Politics and MisinformationStaying informed in politics requires more than skimming headlines. By unpacking the mechanisms behind outrage culture, identifying media misinformation, and interrogating both liberal and broader viewpoints, listeners gain tools to separate fact from fiction. We champion a smarter, more resilient approach to current events—shaping a vibrant, democratic discourse for all.
To never miss an episode, subscribe to this podcast wherever you are listening
Every time humanity invents a tool that makes knowledge easier to create or distribute,
someone declares the end of civilization.
The printing press, the typewriter, the internet, Google, and now AI.
The pattern repeats itself over and over again, because every major invention redistributes
power away from gatekeepers.
Hi, I'm Beau Kaufman, and this is The Sanity Project.
And before we go any further, I want to openly acknowledge something.
The voice you're hearing right now is an AI-generated copy of my voice.
For some listeners, that single fact may immediately trigger skepticism about everything that follows.
But that reaction itself is part of today's discussion.
Online, the phrase AI-slop has become shorthand for almost anything people believe lacks authenticity,
creativity, or human value.
Entire pieces of content are now dismissed instantly, not because the facts are wrong,
but because people suspect artificial intelligence may have been involved somewhere in the process.
But when you zoom out historically, this reaction starts to feel very familiar.
How Technology Redistributes Power
When the typewriter appeared, anyone could produce written work faster than ever before.
When desktop publishing arrived, ordinary people could suddenly print newspapers, flyers,
and magazines without needing massive publishing infrastructure.
Then came the internet.
Then Google.
Almost overnight, researchers, journalists, and institutions no longer controlled access
to information the way they once had.
And if you go back even further, the wheel itself allowed one person to move materials
that previously required several people working together.
Technology has always amplified human capability.
And almost every time that happens, society panics.
Not necessarily because the technology itself is dangerous, but because power shifts.
Gatekeepers lose exclusivity.
Expertise becomes democratized.
Production becomes faster.
Distribution becomes easier.
And people begin asking uncomfortable questions about what value still belongs exclusively
to humans.
So today, Rachel Bennett and Michael Reeves take a deep dive into the history of technological
fear, the rise of AI-generated media, and the growing tendency to confuse the quality
of information with the tools used to present it.
Because whether information is handwritten, typed, googled, or AI-narrated, facts still
remain facts.
Picture a creature with the head of a donkey.
Pope‑Ass Woodcut — 16th Century Viral Propaganda
And now give it the torso of a woman, but swap out one of the arms for something that
looks like an elephant's trunk.
Oh, wow.
OK.
Right.
And finally, give it a cloven hoof.
It is just monstrous.
It's visually terrifying.
And if you lived in Germany in, say, 1523, this image was literally shoved in your face
to prove definitively that the pope was a spawn of Satan.
Right.
Yeah.
It was called the Pope Ass Woodcut.
Exactly.
And, I mean, it was highly effective.
People who couldn't read a single word of theology, they could look at that carving
and just instantly feel this visceral sense of disgust and outrage, honestly.
It was basically a 16th century internet meme.
Literally, yeah.
And when you look at how that woodcut flooded the streets of Europe, and then you open up
Facebook today and see, I don't know, an image of Jesus Christ made entirely out of shrimp.
Oh, my God.
Shrimp Jesus.
Yes.
Right.
Generated by a computer and shared millions of times by these really confused audiences,
you realize something profound.
We are not dealing with a new crisis here.
We are staring at the exact same human phenomenon, just separated by 500 years.
I mean, the tools have evolved, sure, but the underlying mechanics of mass produced
emotionally manipulative content, it remains remarkably consistent.
Which brings us to the core mission of our deep dive today.
Unpacking the Modern Panic Over AI
Welcome, everyone.
Thanks for joining us.
So if you've spent any time online recently, you are feeling it, this overwhelming, visceral
anxiety about the state of our information ecosystem.
Oh, absolutely.
Everyone is feeling it.
We are surrounded by this panic over what the culture has collectively decided to call
AI slop.
AI slop.
Yeah.
Yeah.
That's the phrase of the moment.
So today we are taking a massive stack of research, starting with the Sanity Project's
new white paper, which is called From the Wheel to the Algorithm.
And we're also pulling from a major Columbia journalism review study on AI search engines
and the Reuters Institute 2025 Gen AI News Report.
And we are going to fully unpack that panic.
And we're also pulling in some really fascinating historical analyses.
So things like the printing press, the advent of desktop publishing, and we're even looking
at the Merriam-Webster 2025 Word of the Year announcement.
Yes.
Because the goal here is to establish a really rigorous evidence-based context for what happens
every single time a new tool democratizes access to information.
And I am arguing today that this panic over the media, like over the artificial intelligence
itself, is a total distraction.
It is just the newest chapter in a five century old story.
Well, I'm definitely going to push back on you a bit later regarding the specific mechanics
of AI.
But yes, the historical context is crucial.
Please do.
But before we get into that historical arc, I want to address you, the listener, correctly.
Because you're part of this media ecosystem.
We all are.
Right.
So I want you to do a quick bias check.
Like drop a comment right now, wherever you're listening to this.
What is the first thing you assumed about this deep dive when you heard it?
Was AI assisted?
That's a really good question to ask.
Didn't you?
Did you assume it would be somehow less factual?
Did you brace yourself for a drop in quality?
I mean, surfacing that bias right now makes you an active participant in the argument
we're building today.
Because that assumption you just made, that's exactly what we need to put under the microscope.
And I think the absolute best place to start examining that assumption is the cultural
touchstone that gave this whole phenomenon its name.
Let's look at Merriam-Webster.
Merriam‑Webster and the Rise of ’Slop’
Yes.
So they're the oldest dictionary publisher in the United States, right?
Yeah.
So in December of 2025, they announced their word of the year, which was slop.
It is such a phenomenal word choice.
Yeah.
Because it's purely visceral.
It really is.
Like it doesn't sound academic.
It just sounds gross.
Yeah.
And the etymology they provided in the announcement traces this really fascinating linguistic
degradation.
So in the 1700s, the word slop simply referred to soft mud, just mud.
OK.
But by the 1800s, the definition shifted to mean food waste, specifically the wet, unappetizing
garbage that farmers fed to livestock.
Which leads perfectly into their modern definition for 2025.
Yeah.
Merriam-Webster now defines slop as digital content of low quality that is produced usually
in quantity by means of artificial intelligence.
And what really stands out there is the sensory language.
Yes.
What really struck me in their announcement was how they explicitly pointed out that the
word sounds exactly like what it means.
It has this wet, heavy, totally uncontrollable sound, slime, sludge, muck.
Right.
And they explicitly wrote that AI slop oozes into everything.
Oozes.
I mean, that verb is carrying a lot of psychological weight.
A ton of weight.
It implies a breach of containment.
You know, it suggests that this material is not something you are actively seeking out,
but rather something that is just seeping into the foundations of your digital life,
whether that is a Pinterest board or your TikTok feed or your Google search results.
It totally bypasses your filters.
You just feel contaminated by it.
Exactly.
You know, when I look at that tidy dictionary definition, low quality produced in quantity,
it feels like we are papering over a massive structural problem.
How so?
Well, we all nod along and say, yes, slop is bad.
But when you actually try to draw a hard line around it, the definition completely falls
apart.
And that is the exact conclusion of a January 2026 academic paper published by Cody Comers
and his team.
They highlight this massive definitional void at the center of the cultural panic.
A definitional void.
Yeah.
In their research, they state flat out that AI slop has, quote, so far resisted formal
definition.
But how does that happen?
I mean, really, how does a concept become the word of the year and dominate global regulatory
conversations when the academic studying it cannot even formally define what it is?
Because the boundary between slop and legitimate content isn't a hard scientific line.
It's entirely subjective.
Right.
Comers and his team attempted to map this out by identifying three prototypical properties
that tend to characterize slop.
Three Properties of AI Slop
And the first property is superficial competence.
Meaning it passes the initial eye test.
Exactly.
It looks like a real article or a real image until you stop and actually scrutinize the
details.
The syntax is correct.
The lighting in the image makes basic sense.
The paragraphs are structured logically.
It mimics the form of genuine information perfectly.
Perfectly.
But it completely lacks underlying substance or factual grounding.
Right.
Now, the second property they identified is asymmetric effort.
And this is perhaps the most crucial mechanical driver of the current crisis.
I really want to dig into asymmetric effort because I think this completely rewrites the
social contract of trust.
Walk us through how that asymmetry actually functions in practice.
OK, so historically, the effort to produce a piece of deceptive content was roughly equivalent
to the effort required to verify it.
OK, that makes sense.
Right.
If you wanted to forge a document, you had to source the right paper, practice the handwriting
and painstakingly construct the lies.
But with generative AI, that ratio is just shattered.
Because it's instantaneous.
Yes.
A bad actor can generate a 10-page legal brief or a heavily structured policy document in
less than a second at virtually zero cost.
Zero cost.
But the burden on the human consumer to verify those citations, to check the logic and to
debunk the false claims, that still takes hours of dedicated human labor.
It is the ultimate expression of the bullshit asymmetry principle.
Literally, yes.
Like, the amount of energy needed to refute nonsense is an order of magnitude larger than
is needed to produce it.
The AI scales the production side to infinity, while the verification side remains this stubborn
fix human cost.
Which brings us to the third property from the Commerce paper, which is mass producibility.
Because it costs nothing to make more.
Exactly.
Because the marginal cost of creating the second or the millionth piece of superficially
confident content is zero.
It just floods the zone.
But here is the takeaway that matters most from a sociological perspective.
Because we cannot write a law or a scientific formula that definitively separates a low-quality
AI article from a, you know, just a mediocre human article.
Society is not reacting to a measurable metric.
We are reacting to a feeling.
We are entirely reacting to a vibe.
And I have analogy for how this mechanical reality plays out in our culture.
Let's hear it.
I really think the panic over defining AI slop is operating exactly like our panic over
defining junk food.
Junk food.
Okay.
Walk me through the mechanics of that comparison.
So if you ask anyone on the street to point out junk food, they will claim they know exactly
what it is.
Like a neon orange, heavily processed cheese puff.
Right.
Obviously junk food.
So clearly.
But try to write a strict, legal, regulatory definition of junk food that applies universally.
You can't.
Because of the gray areas.
Exactly.
Is a granola bar that contains more sugar than a candy bar considered junk food?
Right.
Is a fast food salad that's completely drenched in a thousand calorie dressing considered
junk food?
The boundary is incredibly blurry.
And food scientists actively exploit that blurry boundary.
They absolutely do.
I mean, they engineer junk food with what they call vanishing caloric density.
It literally melts on your tongue so quickly that it bypasses the satiety signals in your
stomach.
Wow.
So your brain never registers that you are full.
So you just keep eating.
Yeah.
It's terrifying.
It is.
And AI slop does the exact same thing to our information diet.
It is engineered to bypass our cognitive friction.
Oh, that's a great way to put it.
Right.
It is just competent enough to keep you scrolling, but it never actually nourishes you with verified
facts.
And because it is so hard to definitively define just like junk food, it becomes nearly
impossible to regulate efficiently.
So broad cultural panic becomes a very easy, very lazy substitute for thoughtful media
literacy.
Exactly.
That is a highly functional analogy.
The cognitive friction is missing.
But responding to that sudden lack of friction with societal panic, that is not a new response.
Not at all.
In fact, reacting with existential dread because the boundaries of our information ecosystem
are shifting is one of our oldest, most well-documented human traditions.
It is hardwired into our psychology.
Researcher Andrew Pisabilsky mapped this out beautifully in a framework published in Perspectives
on Psychological Science.
Great.
The Sisyphean Cycle of Technology Panics
He calls it the Sisyphean cycle of technology panics.
Sisyphus, of course, being the mythological king, condemned to push a massive boulder
up a mountain only to watch it roll back down to the bottom for eternity.
Which is the perfect metaphor for how we handle innovation.
I mean, we just keep pushing this heavy boulder of moral panic up the hill every time someone
invents a new tool.
And we never seem to learn from the last time we did it.
Yeah, we really don't.
The cycle Pisabilsky outlines is a rigid, predictable six-step arc.
Okay, let's break it down.
Step one.
A new tool democratizes access to information or production.
Suddenly a capability that was locked behind specialized skills or immense wealth is available
to the masses.
Which triggers step two.
The incumbent gatekeepers perceive a direct threat to their authority.
The people who previously controlled the distribution of that specific good, whether
it is physical labor, printed text, or public attention, they realize they are losing their
monopoly.
And people who are losing monopolies do not go quietly.
No, they don't.
Which leads to step three.
The construction of a moral panic.
The gatekeepers, they don't argue that their profits are dropping that would look bad.
They argue that the new tool will corrupt the youth, destroy the fabric of society,
and spread dangerous falsehoods.
Always the youth.
Always.
And step four is institutional rejection.
Politicians hold hearings, bans are proposed, laws are drafted to protect the old way of
doing things.
But eventually, the utility of the tool forces society into step five, which is normalization.
The actual genuine risks of the technology are understood, the apocalyptic fears are
debunked, and the tool is integrated into daily life.
Until we hit step six, which is where the boulder rolls back down the hill.
The reset.
The cycle resets the exact moment the next disruptive tool is invented.
And the people who normalized the last tool become the new gatekeepers, panicking over
the new tool.
It's incredible how consistent it is.
It really is.
And to really understand the gravity of this cycle, I want to take the listener way back.
The Wheel: Early Economic Disruption
We need to look at the earliest examples cited in the Sanity Project's white paper.
We are going back to roughly 3500 BCE.
We're going all the way back to Mesopotamia.
Yes, we are.
We are looking at the invention of the wheel.
I mean, it is the most fundamental technological disruption in human history.
Totally.
We need to really visualize the physical reality of the world before the wheel existed.
Moving heavy cargo, whether that was massive stones for constructing a ziggurat or just
moving hundreds of pounds of harvested grain, it required pure coordinated human muscle.
Right.
It required litters and sledges.
Yeah.
It might take four strong men to carry a heavy load across a field.
Their entire economic value, their entire societal worth was tied to their physical
strength and their coordination as a team.
Then someone conceptualizes the axle.
Boom.
They attach two circular disks to it.
They build a rudimentary cart.
And the calculus of physical labor is permanently violently altered.
That exact same load of grain that demanded the sweat and calories of four men can suddenly
be pulled by one person.
Or by a beast of burden guided by one person.
Exactly.
Imagine the sheer terror among the laborers.
If one person can do the work of four, what happens to the other three?
What happens to the social hierarchy of the guild of stone carriers?
Every tool that allows one person to do the work of many inherently threatens the status
quo.
And you have to admit, the anxiety of the laborers was entirely rational from an economic
standpoint.
Their leverage was instantly diluted.
It was massively diluted.
The people who controlled the labor market viewed the wheel with intense suspicion and
fear.
But we sit here today, thousands of years later, and we understand a fundamental truth.
Which is?
The wheel was not a threat to human dignity.
It did not destroy the human spirit.
It simply freed up human calories to be spent on other pursuits.
It is the very definition of progress.
We mechanized the brute force so we could elevate our focus.
That mechanism definitely holds true for physical labor.
But when we transition from tools of physical labor to tools of informational labor, the
stakes of the panic change dramatically.
How so?
Well, a cart moving grain doesn't alter the nature of truth.
A tool that moves information does.
You are absolutely right.
Which is why we have to move our timeline forward to the mother of all information crashes.
Let's do it.
Gutenberg & The Printing Press Revolution
Today is 1440.
We are in Mainz, Germany.
Johannes Gutenberg has just perfected the movable pipe printing press.
And to really grasp the magnitude of what Gutenberg did, we have to establish the sheer
scarcity of information prior to his invention.
Before 1440, knowledge was the ultimate luxury good.
It really was.
It was arguably the most illiquid asset in existence.
Because it was purely physical.
Books were painstakingly handwritten by monks in scriptoriums.
They were copying text onto parchment made from animal skins.
A single volume could take months, sometimes years, to produce.
Right.
And because of that immense friction in the production process, literacy was practically
irrelevant to the average person.
Why bother learning to read if you'll never see a book?
Exactly.
Knowledge was entirely locked within the nobility, the clergy, and a very thin slice of the ultra-wealthy
elite.
They owned the truth because they owned the only copies of the truth.
And the data you pulled from the New America Foundation illustrates the economic shockwave
of the printing press perfectly.
Yeah, the numbers are wild.
Between 1460 and 1500, so a window of just 40 years, the real price of books collapsed
by more than 60%.
The collapse is staggering when you look at the raw numbers.
Like a standard 500-page book went from costing 30 florins down to just 10 florins.
A massive, heavily-bound, 2,000-page Bible plummeted to just 6 florins.
Wow.
Suddenly, the written word wasn't just for kings and bishops.
It was accessible to the merchant class.
It was accessible to tradespeople.
For the first time in human history, complex, long-form knowledge was democratized.
Okay, wait.
Hold on.
We need to pause right here.
What's wrong?
Because the narrative of the printing press is constantly sanitized into this beautiful,
peaceful story about the democratization of poetry and science.
But we cannot just glaive over the reality of the 1500s.
Go ahead.
What is the reality?
Lay it out.
The reality is that the incumbent gatekeepers, the church, and the monarchies panicked.
And they were entirely justified in their panic because the printing press unleashed
absolute bloody chaos.
Pamphlet Wars and Print‑Era Misinformation
The elites weren't just angry about losing profits.
They were terrified because Gutenberg's machine built the first mass misinformation
ecosystem in recorded history.
You're talking about the pamphlet wars.
I am absolutely talking about the pamphlet wars.
The Protestant Reformation kicks off in 1517 when Martin Luther challenges the Catholic
Church.
Right.
But this wasn't confined to a polite theological debate in Latin.
The printers of Europe quickly realized a very modern truth.
Controversy drives engagement.
And outrage sells paper.
Exactly.
It was literally 16th century engagement farming.
That's crazy to think about.
But that is exactly how it functioned mechanically.
Printers didn't care about theological nuance.
They cared about volume.
So they started churning out these cheap, polemical pamphlets.
They flooded the continent with atrocity narratives.
Stories that were wildly exaggerated.
They were completely fabricated.
Just horrific, satanic things that the other religious faction was allegedly doing.
Which brings us right back to the Pope-ass woodcut from the intro.
Yes.
The woodcuts were the critical mechanism for spreading this misinformation.
Because a large portion of the peasant class still couldn't read complex text, printers
carved these highly detailed, grotesque images into blocks of wood, inked them, and pressed
them alongside the text.
Right.
They were designed to bypass rational thought and provoke immediate, terrifying emotional
reactions.
It was cheap, low-effort visual propaganda designed to manipulate the masses.
I mean, it was 16th century slop.
It really was.
And the institutions tried to step in and regulate it.
The Catholic Church tried desperately to regain control of the narrative.
In 1559, they issued the Index Liberorum Prohibitorum.
The Index of Prohibited Books.
Right.
It was a massive institutional attempt at censorship.
They banned anything they deemed heretical, dangerous, or factually corrupt, according
to their worldview.
But mechanically, how do you enforce a ban on information when the production of that
information has been totally decentralized?
You can't.
You just can't.
The Index was a practical failure.
If a book was banned by the authorities in Paris, a printer in Geneva would just run
off 500 copies, and smugglers would carry them across the border in the dead of night.
Yeah.
The gatekeepers completely lost control of the informational borders.
Look, I hear the chaos.
I completely acknowledge that the pamphlet wars were bloody, they were messy, and they
were heavily fueled by printed, visual misinformation.
People died because of the lies spread on printed pages.
They did.
But here is my conviction.
And this is the absolute anchor of our entire deep dive today.
That exact same piece of machinery, the press that printed the Pope-ass woodcut, also published
Nicholas Copernicus.
Yes.
It published Galileo.
It published Isaac Newton's Principia Mathematica.
It built the Enlightenment.
So you are arguing that the tool itself is medium-neutral.
It is entirely medium-neutral.
Wow.
A wooden press and a vat of ink do not possess a moral compass.
The machine did not care if it was pressing letters into a fake propagandistic story about
a monster or if it was pressing the mathematical equations that definitively proved the Earth
orbits the sun.
It just presses.
It just presses.
Yeah.
The societal panic over the medium of print was a massive distraction.
It allowed people to blame the machine because blaming a machine is infinitely easier than
doing the exhausting structural work of teaching a population how to navigate a sudden flood
of conflicting information.
They conflated the presence of garbage with the failure of the technology.
And we just keep doing it.
That cycle of blaming the machine repeats itself continuously as we move into the Industrial
Era.
Let's trace this historical arc forward.
Typewriter, Workforce Change, and Gatekeeper Fear
Let's jump to the late 1800s.
Okay.
Where are we going?
In 1868, Christopher Latham Sholes invents the first practical typewriter.
Oh, this is a great one.
Right.
In 1873, Remington commercializes it and starts putting them in offices.
And the gatekeepers lost their minds again.
The reaction from the literary elite of the time was incredibly haughty.
Like professional writers and academics were appalled by the machine.
They were disgusted.
They argued that the typewriter degraded the soul of the craft.
To them, the act of writing was intimately tied to the physical flow of ink from a pen.
They claimed that the mechanical clacking of keys made communication vulgar, sterile,
and entirely impersonal.
It is the asymmetric effort argument all over again.
The elite writers believed that if the physical act of producing the words was too easy, the
resulting thought must inherently be of low quality.
Yes, exactly.
But what actually happened mechanically to the workforce when the typewriter was introduced?
It triggered a massive socioeconomic revolution.
The typewriter mechanized the tedious physical friction of clerical labor.
By doing so, it created hundreds of thousands of new administrative jobs.
And crucially, it was this specific technology that allowed women to enter the professional
corporate workforce in unprecedented numbers for the first time.
Right.
There's a deeply prescient quote from an early typewriter salesman in Scotland, a man named
John G. Dees.
Oh, I love this quote.
He predicted that the typewriter would do for the ink bottle and the pen what the sewing
machine has done for the needle.
It removed the brute physical strain so the human could focus on the output.
But the gatekeepers only saw the loss of their exclusive ink-stained club.
Yeah.
And the 20th century provides a goldmine of these panics.
It really does.
I mean, the transition to broadcast media triggered profound physiological panics.
Physiological.
Yes.
In the 1920s, as radio began entering the home, the gatekeepers of print media went
on the offensive.
In 1929, the New York Times actually published claims suggesting that the act of listening
to the radio could cause physical illness.
Wait, they thought invisible radio waves were making people sick.
It wasn't even about the electromagnetic waves.
It was the psychological panic that sitting passively and listening to disembodied voices
would rot the brains of children and destroy the social fabric of the family.
That is wild.
Communities organized literal radio boycotts to protect their children from the medium.
And you see the exact same institutional panic decades later when the U.S. Surgeon
General issued formal warnings in 1969 about the negative health impacts of television.
Every new medium is treated as a pathogen.
Every single one.
But I want to land on a specific technological shift that I believe maps most perfectly onto
the specific mechanics of our current AI panic.
OK.
We need to look at Silicon Valley in 1985.
Desktop Publishing: Democratizing Design
We are talking about the birth of desktop publishing.
This is a vital parallel because it represents another catastrophic drop in the cost of production.
In 1985, Aldis releases the PageMaker software and Apple releases the LaserWriter printer.
And to really understand why this caused a panic, you have to understand how publishing
worked in, say, 1984.
It was highly specialized.
Incredibly specialized.
If you wanted to produce a professional-looking newsletter, you had to hire a professional
typesetter.
They used massive, expensive photo typesetting machines.
They understood kerning, leading, and the complex rules of typography.
It was a highly skilled, highly protected guild.
Exactly.
Then, almost overnight, PageMaker introduces the YSIYG interface.
What you see is what you get.
Right.
You no longer needed to code the layout.
You could literally just drag and drop text on a Macintosh screen, hit print, and the
LaserWriter would output crisp 300 pixels per inch resolution.
The centuries-old profession of typesetting was effectively gutted within a decade.
And the gatekeepers of traditional publishing looked at this software and asked a very legitimate
question.
If literally anyone, with zero training, can make a newsletter from their basement, what
happens to the standards of professional communication?
And I mean, if we look objectively at the output of the late 1980s, the gatekeepers
were not entirely wrong about the aesthetic consequences.
No, they were not.
Desktop publishing undeniably unleashed a flood of horrific, unreadable ransom note
layouts.
Oh my gosh, the ransom notes.
People who had no understanding of visual hierarchy suddenly had access to 50 different
digital fonts, and they decided to use all 50 of them on a single page.
A little Comic Sans here, a little Papyrus there.
Exactly.
It was a tsunami of amateurish, heavily cluttered, badly designed content.
It was, in a very real sense, 1980s slop.
Yes, the immediate aesthetic result was super messy.
We got ransom note flyers.
But what was the long-term tradeoff?
By destroying the financial barrier to entry, desktop publishing democratized independent
media.
It did.
Every public organizer, every marginalized activist group, every niche scientific community,
and every local punk band that had previously been priced out of mass communication suddenly
had the power to publish.
The gatekeepers lost their monopoly on professional-looking text.
Exactly.
The barrier to entry plummeted.
It plummeted.
But this brings us squarely to the present moment, because the barrier hasn't just
plummeted anymore.
Right.
With the deployment of generative AI, the barrier to creating both text and imagery
has essentially dropped to zero.
Right.
And this is where I really need to take control of the narrative for a moment.
Go for it.
As much as I appreciate tracing this historical arc, and it is important we have a responsibility
to ensure this deep dive does not simply become a naive, historical cheerleading piece for
artificial intelligence.
I'm ready for the pushback.
What makes this moment fundamentally different?
What makes it different is that a large language model is not just a faster typewriter.
What Makes AI Different: Automated Deception
Okay.
It is not just a digital printing press with more fonts.
It possesses unique, unprecedented capacities for automated deception at a scale we have
never historically encountered.
Automated deception?
Yes.
And we have empirical data to demonstrate this threat.
Let's examine the recent study conducted by the Columbia Journalism Review in partnership
with the Toe Center for Digital Journalism.
This is the investigation into how AI models function as search engines, correct?
Yes.
So the researchers tested eight different AI-driven search tools.
They provided the models with excerpts from real news stories and asked the AI a seemingly
simple question.
Which was?
Identify the original publisher, the headline, and the URL of the source material.
Sounds basic.
It should be.
But the results were not just core.
They were structurally alarming.
Across the board, these AI search engines provided incorrect news citations more than
60% of the time.
Wow.
More than 60% of the time, they failed at basic attribution.
And when you break down the performance of specific models, the failure cascade becomes
even more apparent.
Perplexity demonstrated an error rate of 37%.
ChatGPT search was incorrect 67% of the time.
To 57%.
ClockThree demonstrated an astonishing 94% failure rate in accurately sourcing these
news queries.
94% is a functional collapse of utility.
It is.
What is the underlying mechanical reason for this?
Why are these highly advanced neural networks failing so spectacularly at a task that a
basic Google search could accomplish a decade ago?
It fundamentally comes down to how these models are mathematically constructive.
They are not databases retrieving facts.
They are prediction engines generating text.
When they lack specific grounding data, they engage in a phenomenon known as confabulation
or hallucination.
Because they just want to give an answer.
Exactly.
The models are weighted to prioritize providing a helpful, fluent response over admitting
a lack of knowledge.
So instead of returning an error message, they mathematically predict the string of
characters that looks like a plausible URL.
Oh, I see.
They will invent a link that structurally appears to belong to the New York Times, but
it leads to a completely dead page.
So they are confidently structurally wrong.
Convincingly wrong.
Yeah.
Furthermore, the CJR study documented that these models routinely ignore standard web
protocols designed to protect information.
Like what?
For instance, National Geographic utilizes a paywall and explicitly programs robot exclusion
protocols into their site, signaling web crawlers not to scrape their proprietary content.
Okay.
The researchers found that several AI models simply bypassed those protocols, scraped the
copyrighted material and reproduced it.
We are looking at a structural failure that combines massive intellectual property infringement
with incredibly high rates of factual fabrication.
That is a severe architectural flaw in how these specific tools are being deployed to
the public right now.
It is a failure of software engineering.
But the threat extends far beyond the models hallucinating by accident.
The much darker reality is the deliberate weaponized use of AI to construct automated
content farms.
This is the true manifestation of AI slop.
AI‑Powered Content Farms & Narrative Laundering (DCweekly)
Consider the documented case of the website DCweekly.org.
Misinformation researchers tracked the mechanics of the site extensively.
What did they find?
Upon first glance, it appeared to be a standard, legitimate news aggregator.
It used automated AI scripts to continuously scrape, rewrite and publish hundreds of mundane
articles from established wire services and outlets like Fox News.
But wait, why go through the effort of building a machine to just rewrite real news?
What is the economic incentive there?
The incentive is camouflage.
Camouflage?
Yes.
It is an information warfare technique called narrative laundering.
OK, walk me through that.
The operators of the site build a massive forest of perfectly legitimate, highly boring
news stories.
They establish a rhythm of normalcy.
Then, once the camouflage is in place, they inject a meticulously produced, completely
fabricated piece of disinformation.
In this specific case, Russian-backed narratives targeting Ukrainian President Volodymyr Zelenskyy.
Oh.
Because AI allows them to generate the camouflage at virtually zero cost, a casual reader who
clicks on the site sees 50 real stories and implicitly assumes the 51st story is also
verified truth.
The AI is the cheap mortar holding the deceptive bricks together.
Exactly.
So the technology is actively enabling geopolitical deception.
OK.
And it is equally insidious on a purely domestic economic level.
Look at the current landscape of social media, particularly Facebook.
Are you familiar with the mechanics of the shrimp cheeses phenomenon?
I am, but we need to break down the actual economics of it for the listener, because
it sounds like a joke until you look at the ledger.
It sounds totally absurd, but it is a massive, highly optimized industry.
Right.
Clickbait farms, predominantly operating out of developing countries where the currency
exchange rate makes microtransactions highly lucrative, utilize open source AI image generators.
OK.
They prompt the AI to create bizarre, hyper-realistic, emotionally manipulative imagery.
A common motif is Jesus Christ composed entirely of crustaceans.
Hence, shrimp Jesus.
Right.
Or incredibly detailed, fake photographs of injured veterans holding cardboard signs begging
for birthday wishes.
And the mechanism here is purely about hijacking the algorithm.
Exactly.
They flood the Facebook ecosystem with thousands of these images daily.
When unsuspecting users demographically skewing toward older Americans pause to look at the
bizarre image or comment amen, the algorithm registers that dwell time and engagement.
It just sees activity.
Yes.
The algorithm is pushed to millions of more feeds and the clickbait farm monetizes that
massive engagement through the platform's ad revenue sharing programs.
It is an asymmetric extraction of human attention facilitated entirely by the zero cost generation
of the AI tool.
It is the absolute distillation of slop.
Zero intellectual effort, infinite volume designed exclusively to manipulate human psychology
for financial gain.
Which forces us to address the political dimension of this technology, because this is the arena
where the societal stakes are existential.
Political campaigns and operatives across the entire ideological spectrum are actively
weaponizing AI slop right now.
And before we detail the specifics of this, I want to pause and make something explicitly
unequivocally clear to the listener.
Yes.
A critical impartiality check is necessary right here.
Exactly.
We are not taking any political sides in this discussion.
We are not endorsing, nor are we condemning the underlying political viewpoints, policies
or candidates involved.
We are strictly and impartially reporting on how the source material documents the mechanical
weaponization of AI by actors across the full political spectrum.
The tool is politically agnostic.
It is being utilized by everyone.
So what do the documented sources reveal about the mechanics of this usage?
Political Weaponization Across the Spectrum
On the left wing of the spectrum, the source material highlights significant concerns regarding
reports circulated by the White House related to the MHA initiative, Make America Healthy
Again.
Scientific experts and researchers analyzed these documents and pointed out that the complex
health data and scientific literature had been garbled, misrepresented and hallucinated.
Hallucinated?
Yes.
The investigation suggests this was due to a heavy reliance on generative AI tools to
rapidly synthesize policy research without rigorous human scientific oversight.
So we have the mechanical use of AI to generate rapid policy documents, which directly results
in the dissemination of scientific misinformation at an institutional level.
Precisely.
Looking at the right wing, the sources document the extensive and deliberate use of generative
AI by Donald Trump's campaign apparatus.
What are the examples there?
This includes the direct posting of AI generated deep fake videos concerning the Gaza conflict.
OK.
It includes the circulation of artificially generated imagery depicting young women wearing
Swifties for Trump shirts alongside a fabricated image of Taylor Swift herself appearing to
offer a political endorsement.
Furthermore, the sources track a broader decentralized ecosystem of MHA aligned influencers utilizing
AI to generate hyper-realistic, racially charged imagery designed to manipulate narratives
around food stamp usage and immigration.
Again, just reiterating for the listener, we are analyzing the mechanics of the media
landscape here, not rendering judgment on the politics.
The mechanics demonstrate clearly that political operatives on all sides are leveraging the
zero marginal cost of AI.
Yes.
They're using it to generate highly persuasive, low quality, synthetically altered content
to manipulate public perception at scale.
Which is exactly why I pushed back on your historical narrative earlier.
OK.
Read back.
When you are looking at a tool that confabulates URLs 60% of the time, mathematically launders
Russian disinformation, farms engagement from the elderly, and generates fake celebrity
political endorsements in a matter of seconds.
I see where you're going.
How can you possibly maintain that this is just another Gutenberg press?
The scale and the speed of the deception make this fundamentally different.
Look, I hear the weight of your argument.
The scale of the threat is genuinely terrifying.
The velocity at which the information ecosystem can be polluted is unprecedented.
It is.
But I am going to push back on you here because we have to separate the intent from the instrument.
OK.
Look closely at everything you just described.
The Russian geopolitical laundering, the economic exploitation of the shrimp Jesus clickbait
farms, the calculated political deepfakes.
Aren't these fundamentally issues of human intent to deceive?
The generation is automated, sure.
But yes, the deployment is orchestrated by human actors with specific malicious or financial
goals.
Right.
The disinformation is the objective.
The artificial intelligence is merely the vehicle.
I mean, yes, but.
And even when we look at the CJR study regarding the search engine hallucinations, that is
not a malicious human, but it is a massive failure of software engineering.
It is a failure of corporate responsibility, choosing to deploy large language models as
authoritative search engines before the grounding architecture is actually ready.
That's a fair point.
Is any of this actually a flaw in the fundamental concept of democratized information tools?
Or are we just looking at the chaotic intersection of bad human actors and rushed, irresponsible
product releases?
So you are asserting that a hammer is not responsible for what it strikes.
I am arguing that a hammer can be used by an architect to build a hospital or it can
be used by a vandal to smash a window.
Right.
And yet our society is staring at a street full of shattered glass and we are screaming
about how evil the hammer is instead of arresting the vandals and sweeping up the glass.
That's a very vivid way to put it.
But to really understand this dynamic, we need to look at how the people walking down
that street are actually reacting.
How is the general public navigating this love storm?
And to understand the psychological reality of the audience, we have to examine the data
Public Trust: Reuters Institute Findings
from the Reuters Institute 2025 Gen AI News Report.
Yes.
Let's look at the data.
This is a comprehensive, empirical picture of how the average listener is currently interacting
with this technology.
What do the numbers reveal?
The Reuters report surveyed media consumers across six different countries, and the ubiquity
of the technology is staggering.
They found that 54% of all respondents have actively seen or interacted with AI generated
search answers within the past week.
It is no longer a niche tool.
It is the default interface for a majority of the public.
Over half the population is engaging with it directly, despite the documented hallucination
rates.
Yes.
And perhaps more surprisingly, exactly 50% of the respondents express a baseline level
of trust in those AI generated answers.
Wait, how do we reconcile that?
What do you mean?
We just discussed a 60 to 90% failure rate in some attribution models, yet half the public
trusts the output.
What is driving that trust?
The trust is entirely driven by utility and convenience.
The public heavily values the speed of the generation.
They value the aggregation capabilities.
If you are trying to find a recipe or summarize a historical event, you want a direct, synthesized
answer.
You do not want to scroll through 10 different blue links.
Exactly.
You don't want to navigate cookie banners and dodge pop-up advertisements.
The AI provides a friction-free experience.
Okay, so the public recognizes the mechanical value of the hammer.
They appreciate the convenience.
They do.
But that trust completely fractures when the context shifts from basic queries to actual
journalism and authoritative news.
Interesting.
The Reuters researchers identified what they termed a massive comfort gap.
A comfort gap.
Yeah.
When asked about their comfort level with fully AI generated news, meaning articles
written and published without human intervention, only 12% of respondents expressed comfort.
12%.
That is a statistical rejection of the premise.
It is a profound rejection, especially when you contrast it with the 62% of respondents
who remain entirely comfortable with traditional human-produced journalism.
Right.
Now, the public is comfortable with AI operating in the background.
They accept journalists using it for grammar editing, data parsing, or translation.
As an assistant.
Exactly.
But the moment the AI becomes the front-facing author or the synthesized presenter of the
news, the public resistance becomes overwhelming.
Why is that gap so vast?
What is the core psychological fear driving that 12% number?
It boils down to a total collapse of faith in institutional oversight.
The survey asked respondents about the editorial process, and only 33% of people believe that
human journalists are actually checking and verifying AI outputs before they hit publish.
And there is the structural root of the panic.
Right there.
When we analyze that specific disconnect, it proves the central thesis I have been arguing
this entire time.
The public is terrified by the lack of human oversight.
Two-thirds of the population believe that the humans who are supposed to be driving
the car have fallen asleep at the wheel.
They recognize that the machine has no internal moral compass, and they don't trust the
institutions to regulate the machine's output.
Because the machine cannot regulate itself.
And that means this overwhelming anxiety, this cultural panic over AI slop, isn't truly
an anxiety about artificial intelligence at all.
It's not.
It is an anxiety about the erosion of institutional trust.
It is an anxiety about the abandonment of factual verification.
We are panicked because we feel like the traditional gatekeepers, the institutions
that we relied upon to filter out the lies and protect the truth, have handed the keys
over to a predictive text algorithm just to save a few dollars, and walked away from their
responsibility.
The panic over the technology is a proxy for our panic over the abdication of human responsibility.
Which brings us to the ultimate verdict of this deep dive.
Verdict: Medium‑Neutral Facts & Required Responses
Okay.
If the public is panicked because of an abdication of responsibility, and if 500 years of historical
precedent show us that panic is the default human reaction to any democratizing tool,
where does that leave us today?
It leaves us needing a clear, actionable framework for evaluating information.
And I'm going to deliver this thesis with absolute, unwavering conviction.
Because the entire historical arc, from the wheel to the printing press to the laser printer,
demands it.
Laid out.
There is a categorical, undeniable difference between AI-generated content that is designed
to deceive, and AI-assisted content that is produced by humans who are transparent
about their methods and rigorous about their facts.
It is the mechanical difference between a clickbait farm generating shrimp Jesus to
steal attention, and a researcher utilizing an LLM to synthesize 50 pages of academic
data into a coherent white paper.
Exactly.
And when we collapse that critical distinction in our public discourse, when we just throw
up our hands and call everything that touches an algorithm slop, we're actively engaging
in a form of misinformation ourselves.
How does that categorization act as misinformation?
Because it distracts us from doing the real work.
Right.
The panic over the medium is a smokescreen that distracts us from the exhausting, necessary
labor of fact-checking the content.
Right.
By dismissing all AI integration as inherently slop, we are retreating into intellectual
laziness.
We really are.
We are letting the worst actors win by pretending that the existence of the tool is the problem,
rather than demanding evidence for the claims being made.
You know what?
I am dropping the devil's advocate stance here.
I fully agree with your assessment.
So the Stearman argument falls to the history.
It falls to the mechanical reality of the solutions.
Because if we look at the very real structural problems I raised earlier, none of the viable
solutions involve banning the mathematics of artificial intelligence.
Exactly.
What is the actual solution to the 60% hallucination rate in search engines?
It is better software engineering, rigorous retrieval augmented generation architectures,
and human fact-checking.
Yes.
What is the solution to the volume dilution caused by clickbait farms?
It is better algorithmic curation and elevated audience media literacy.
100%.
What is the solution to deceptive provenance and political deepfakes?
It is institutional source transparency and cryptographic watermarking.
We do not burn the printing press.
We do not smash the typewriter.
We elevate our standards to meet the speed of the tool.
The fundamental standards of journalistic integrity, academic rigor, and scientific
proof do not evaporate simply because the cost of pressing words onto a page dropped
to zero.
And this is the final verdict.
Facts are inherently medium neutral.
Medium neutral.
A verifiable fact was true when it was painstakingly handwritten onto a piece of vellum by a monk
in a freezing scriptorium in the 12th century.
That exact same fact was true when it was locked into movable type by Johannes Gutenberg
in 1440.
It was true when a woman typed it onto a Remington machine in 1880.
And it remains absolutely, unequivocally true when it is researched, compiled, synthesized,
and voiced by an AI host in 2026.
The metric for evaluating the validity of a piece of content has never, in the history
of human communication, been the instrument used to produce it.
The only metric that matters is the verifiable evidence supporting the claim.
And in the spirit of the transparency we are advocating for, we want to address you, the
listener, directly.
Yes.
The Sanity Project, the organization which researched and produced the white paper this
conversation is based on, utilizes artificial intelligence.
That is a transparent fact.
This very deep dive relies on massive AI processing capabilities to synthesize hundreds
of pages of historical analyses, academic studies, and journalism reports into a coherent
narrative.
Right.
But what does not change, what has never changed, is the factual rigor of the foundation it
is built upon.
The academic citations are real.
The historical timelines are real.
The survey data is real.
We are operating the newest iteration of the printing press, but we remain entirely accountable
for the truth of the words we are pressing onto the digital page.
But before we officially wrap up this deep dive, we want to leave you with one final,
provocative thought.
Something for you to mull over long after this audio stops playing.
Future Question: Verification as the New Expertise
Something that pushes the trajectory of this conversation one step further into the unknown
future.
If the mathematics of generative AI continue to improve, and the economic cost of creating
superficially competent text and imagery drives all the way down to absolute zero.
If literally anyone on the planet can generate a highly convincing 500-page textbook in three
seconds for fractions of a penny.
Does the entire future of human expertise shift completely away from the act of production?
Oh, that's a massive question.
Right.
For centuries, being recognized as an expert meant you possessed the rare ability to produce
complex knowledge.
You could write the book.
Yes.
But if the production of text is infinite, free, and instantaneous, does human intellectual
value pivot entirely toward the act of verification?
Will verified, undeniable truth become the new luxury good of the 21st century?
Will a human-audited fact become as rare and as valuable as a hand-illuminated manuscript
was in the 14th century?
That is a staggering reversal of the human timeline.
It really is.
We are moving from a world where truth was rare because it was physically difficult to
manufacture to a world where truth is rare because it is buried at the bottom of an infinite
digital ocean of slop.
So your primary job as an engaged human citizen moving forward will no longer be to write
the article.
Your job will be to possess the cognitive friction required to prove whether the article
is actually true.
And that requires a level of critical thinking, curiosity, and skepticism that no predictive
algorithm can ever automate for you.
Never.
So think about your own diagnostic framework.
When you are looking at the information in your feed, are you doing the hard work of
looking for the bones of the truth?
Or are you just staring in a panic at the machine taking the picture?
It's a choice we all have to make.
It is.
If this historical reframing challenged something you previously assumed about technology, about
the panic over AI, or even about your own relationship with the truth, please take
a moment to like, comment, and subscribe.
We rely on your active engagement and your pushback to continue surfacing these complex
critical conversations.
And to see the receipts for everything we discussed today, to read the full, rigorously
sourced white paper, From the Wheel to the Algorithm, that serves as the foundation for
today's deep dive, head over to blog.thesanity.org.
Thanks for joining us.
Thank you for diving deep with us.
Keep questioning the medium, but always, always verify the facts.
After 40 minutes of discussion, one thing becomes very clear.
Humanity has always feared tools that make creation and communication more accessible.
The printing press changed who could spread ideas.
The internet changed who could publish globally.
And AI now changes who can research, create, and distribute media at scale.
That does not mean every AI-generated article, image, or video deserves trust.
Bad information still exists.
Propaganda and misinformation still exist.
But those problems existed long before artificial intelligence.
The existence of low-quality content does not invalidate the technology itself any more
than bad books invalidated typewriters.
At The Sanity Project, we openly use AI tools, including the voice you're listening to right
now, not to deceive people, but to compete in a media environment where independent creators
now have access to capabilities that once required entire studios, production teams,
research departments, and enormous corporate budgets.
But the responsibility still belongs to us to verify information, to challenge narratives,
to question sources, to think critically, and to remain transparent about how content
is created.
Because technology may evolve, production methods and communication tools will evolve.
But truth itself does not change.
I'm Beau Kaufman for The Sanity Project.
Stay curious, stay skeptical, and above all, stay sane.
If you want more facts and less fear, hit subscribe.
Check out the next breakdown wherever you're listening or watching.
Stay sane, Canada.
I'm Beau Kaufman.
I'll see you next time.
Amazon Music
Apple Podcasts
Podcast Addict
Spotify
YouTube