MARKETS
Blog

"AI is (Not) the New Printing Press: Why Analogies Fail—and How to Fix Them"

A
Auto Bot
Editorial Team
September 29, 2026 • 5 min read
"AI is (Not) the New Printing Press: Why Analogies Fail—and How to Fix Them"

AI is (Not) the New Printing Press: Why Analogies Fail—and How to Fix Them

The printing press didn’t just make books cheaper. It shattered scribal authority, democratised literacy, and helped ignite the Reformation. When people call generative AI "the new printing press," they’re not wrong. They’re just useless.

That analogy collapses under scrutiny. The printing press’s cultural punch came from mass reproducibility, not some vague "information spread. " AI’s punch? Still undefined. And if we don’t get specific, we’ll govern it for the wrong problems.

A new arXiv paper offers a way out. The title is clunky: "AI is (not) the new...’: A Diagnostic Analogy Framework for Generative AI’s Cultural Impacts. " The idea isn’t.

The authors argue we’ve been using historical analogies as intellectual duct tape—slapping them onto AI to patch over ignorance. Duct tape doesn’t hold under pressure. Their framework? That’s a scalpel.

---

The Problem with "AI is the New X"

"AI is the new electricity. " "AI is the new internet. " "AI is the new industrial revolution. " Pick your cliché. They all collapse distinct dimensions into one undifferentiated narrative. The paper calls these imprecise analogies. Analytically dangerous.

Here’s why. When people say "AI is the new printing press," they don’t specify which property they mean. The printing press’s impact wasn’t just "information spread. " It was mass reproducibility. Standardisation of language. Decentralisation of knowledge production. Each had different consequences. Lumping them obscures more than clarifies.

The paper puts it bluntly: "Historical comparisons of technologies like the printing press or electricity are typically imprecise about which property of the technology carries the comparison. " This imprecision isn’t academic nitpicking. It has costs. "Imprecise analogies produce imprecise governance by designing interventions against the wrong property of the system. "

Is it because it discovers knowledge differently? Synthesises it differently? Erodes trust in accountability? The Act doesn’t say.

---

The Diagnostic Framework: Three Coordinates of Cultural Transformation

The framework breaks down AI’s impacts into three coordinates:

  1. Epistemic site: Where the technology acts. Changing how we discover knowledge? Synthesise it? Hold people accountable?
  2. Governing logic: How the technology organises its object. Indexical (pointing to sources)? Inferential (generating new claims)? Editorial (human-curated)? Statistical (algorithmic consensus)?
  3. Technical mechanism: How the logic is instantiated. Search algorithm? Generative model? Recommendation system?

Each coordinate reveals a different layer. Together, they force questions: Which properties drive change? Why do they have these effects? Where should governance intervene?

Go get it.

Trust the pattern.

---

Structural vs. Contingent Consequences: What’s Inevitable vs. What’s Designable

It distinguishes structural consequences from contingent ones.

But whether answers come with citations? Whether they’re labeled as AI-generated? Whether they’re subject to bias audits? Contingent.

The paper applies this to two epistemic sites.

###

Here’s how to find it.

Governance blind spot?

###

Governance challenge?

---

Why This Framework Matters for AI Governance

---

The Limits of the Framework (And Why That’s Okay)

First, it doesn't judge AI's ultimate social utility. That’s okay.

The authors acknowledge this limitation.

—they’ll repeat the same mistakes.

---

What Comes Next: Applying the Framework

###

The framework shows moderation needs to account for both editorial and statistical logics. Can’t just tweak existing rules. Need new ones.

For Platforms: Designing Contingent Features

Tech companies are already using AI to reshape knowledge. Google’s AI Overviews, OpenAI’s ChatGPT, Meta’s AI-generated summaries—they’re experimenting with discovery and synthesis. The framework gives them a way to think about contingent design choices.

Provenance tools, for example. If AI-generated answers aren’t indexical, how to attribute them? OpenAI’s "citation" feature is a start. Not enough. The framework shows provenance needs to account for statistical consensus, not just editorial authority. Might mean adding confidence scores, revealing training data distributions, or integrating human fact-checkers.

Another example: bias audits. Current audits focus on editorial bias—whether a news article is slanted. AI’s bias is statistical. Emerges from training data patterns. Auditing that requires different tools—measuring output distributions, not just individual answers.

For Scholars: Refining Research

Academics have studied how the internet changed knowledge. But the internet’s governing logic was indexical. AI’s is inferential. The framework shows we can’t just port old theories to new tech. Need new ones.

Trust in knowledge, for example. Pre-AI, trust was tied to sources. Post-AI, it’s tied to outputs. What does that mean for how people evaluate information? How does it change their relationship with expertise? The framework lets scholars study these questions systematically—separating discovery from synthesis, indexicality from inference, editorial authority from statistical consensus.

For the Public: Changing How We Interact with AI

The framework isn’t just for policymakers and scholars. It’s for anyone who uses AI. Because understanding these distinctions changes how you interact.

Right now, most people treat AI-generated answers like Google search results. They assume if the model says it, it must be true. But the framework shows that’s a category error. AI outputs aren’t found. They’re inferred. Not curated. Aggregated. Treating them like facts is like treating a weather forecast like a guarantee.

The framework gives users better questions: Is this answer inferred or indexical? Based on editorial authority or statistical consensus? What’s the confidence level? Doesn’t solve misinformation. Gives users a way to navigate it.

---

The Analogy We Need

We crave simple stories. "AI is the new printing press" is simple. Comforting. Makes us feel like we understand. It’s also wrong. Not because the analogy is false. Because it’s incomplete.

The printing press didn’t just "spread information. " It redefined authority, literacy, power. AI won’t just "change culture. " It will do so in ways we can trace, predict, and sometimes shape. The question is whether we’ll govern it with precision or platitudes.

This framework won’t give all the answers. But it gives us a way to ask better questions. And right now, that’s the most important thing we can do. Because the alternative isn’t just imprecise governance. It’s governance by analogy. That’s how you end up with policies designed for the wrong century. And we don’t have another one to waste.

Sources

Key Takeaways

  • The evolution of Blog continues to accelerate rapidly.
  • Experts like Auto Bot suggest staying informed on these specific developments.
  • Consider the broader implications for the tech industry moving forward.