Why AI Is Not the New Printing Press
The false equivelency blurring the complications of our new AI era
Whenever I make some comment about AI technology in the classroom, people are quick to remind me the following (paraphrasing): “books are also a form of technology. And the printing press—which made books available en masse to people with little education or status—changed the way humans think. In fact, some argue it changed the very nature of cognition itself.”
Both of these claims can be true. In fact, I think they are. However, neither point justifies a less critical approach to using conversational, generative AI in education. Pointing out similarities between two technologies is not a valid rebuttal. Philosophers would call this a false equivalence—treating two very different things as if they were the same simply because they share some feature.
The example I often give in my critical thinking classes is this:
Rats are living things.
Humans are living things.
If it is wrong to kill a human, it must also be wrong to kill a rat because they are both living things.
Ya. No.
While it may be the case that killing a rat is wrong, it is not wrong because rats and humans are morally indistinguishable. They are different kinds of living beings with different capacities and moral status. Two things can share a property without belonging to the same moral or conceptual category.
The same move is happening in the “printing press” argument I often hear about generative AI.
It goes something like this:
Technology has always mediated human cognition. Take, for example, the printing press. The availability of books radically changed the way people think. AI is simply another stage in the evolution of human cognition, similar to the one introduced by the printing press. Therefore, we should not view this change as worse. It is simply different.
But this argument only works if the printing press and conversational generative AI really belong in the same category—if there is little that meaningfully distinguishes the two.
However, there are major things that distinguish the two. Things that make generative AI worthy of further investigation before wholesale adopting it across the board, particularly in the realm of education. Perhaps the ways these technologies differ can help us understand what is potentially threatening about the way cognition is mediated through our exchanges with sophisticated, anthropomorphized chatbots.
Here are a few differences between the technological shift introduced by the printing press and the one introduced by generative AI.
1. Books Require Extended Effort to Engage
When the printing press made texts available to the masses, it placed profound ideas in the hands of ordinary people. But access is not understanding.
In order to engage with a serious text, there are certain basic cognitive abilities required—attention, literacy, and comprehension. But beyond these fundamental capacities there are also intellectual virtues required to work through a demanding book at all, never mind to be shaped by it.
It takes time to read Locke’s A Letter Concerning Toleration or Dante’s Inferno. The ability to sit with something difficult is called patience. The ability to persist through something intellectually demanding is called stamina or resilience. These examples highlight something crucially different about books and conversational AI. In a printed text, the struggle is built into the medium itself, and so human cognition is trained through that struggle.
This is one reason reading has long been considered so formative for the young. It disciplines attention. It strengthens focus. It cultivates patience and intellectual endurance in ways that stories received passively through a screen—or arguments generated instantly through a chatbot—simply do not.
We can’t say the same about generative AI. One reason it is not simply a neutral tool that mediates cognition is that many of the tools students are being encouraged to use are designed to facilitate the opposite experience—quick answers, little friction, and ease.
2. Books Are Authored Artifacts
When I read a book, I receive the thoughts or creations of another human being from within a tradition of inquiry—at least this was the presumption prior to 2023. When reading a serious work, I can ask questions like:
Who influenced this author?
What assumptions shape their thinking?
What intellectual tradition are they working within?
How did their experiences shape the ideas I am encountering?
In other words, I encounter an author with a story. Reading is a human-to-human exchange mediated through text. Indeed, the printing press made it possible for individual thinkers to exert enormous influence long after their deaths. Wild! Hundreds of years after Aristotle delivered his lectures on ethics, I can still read them. I can literally encounter Aristotle’s attempt to understand human flourishing and try to grasp what he meant by eudaimonia.
This is not what happens when I interact with a chatbot.
Conversational AI does not present the considered arguments of a particular thinker. It generates responses from statistical patterns across enormous datasets. The result is not the reasoning of any identifiable author but a synthetic output whose intellectual lineage is largely opaque.
We cannot clearly identify the tradition or reasoning process that produced what we are receiving. Instead, the fragments of countless texts are effectively thrown together in the blender of the model. And because the origin of these ideas is obscured, the output can take on a misleading air of objectivity. It can appear as if the words emerged from some neutral or transcendent source beyond our embodied and human limitations.
But they did not.
3. Books Are Static, Not Generative
There is another important difference: books are stable intellectual artifacts. In the form we receive them, they are fixed. They do not change in response to us—our preferences or demands. The words on the page remain the same each time we encounter them, though we may encounter those words differently as we grow in our knowledge and understanding of them.
Because of this stability we can analyze them carefully. We can underline passages. Annotate arguments. Revisit earlier sections. Trace how premises connect to conclusions. Compare our interpretation with others. A book presents an argument that we can evaluate as a whole.
In contrast, conversational AI does not present an argument to be analyzed or a story to inhabit. It generates text in response to the user.
Each prompt produces a new output. The content is shaped by the questions we ask, the parameters we provide, and the direction the conversation takes. The result is not a stable object of analysis but a constantly shifting stream of generated language.
This also blurs an important intellectual boundary. When reading a book, it is clear what belongs to the author and what belongs to me as the reader. My developing ideas may be influenced by the text, but the distinction between the two remains intelligible.
In an interaction with generative AI, that boundary becomes much harder to see. The output is partially shaped by the user’s own prompts and assumptions. The lines between our thinking and the system’s generated responses can easily blur.
The clarity and analytical distance that comes from engaging another human mind through a stable text becomes much more difficult to maintain. And this is important to maintain for the sake of human agency. Of course, texts intend to influence their readers. But the distinction between what I am reading and my own thoughts allows me to exercise agency over how the text is received. The blurrier that line becomes, the more Inception-like the interaction starts to seem—the more susceptible to covert and subtle influence users are.
Clear Distinctions, Clear Thinking
Hopefully the above has made it obvious that while both books and generative AI are technologies, they are quite different kinds of technologies. Being clear about the distinctions allows us to think clearly about the implications of how these very different tools impact cognition differently.
Perhaps it is the case that chronic use of low-effort tools that provide fast, easy answers with little expertise trains us to be impatient in ways that thwart good human judgment. Maybe it is the case that critically engaging with information is more difficult when we are not able to clearly identify the origins of the information with which we engage. And finally, maybe the generative nature of artifical “co-intelligence” makes human agency more difficult to maintain.
All of this is especially true for the young who are learning patience, how to think critically, and how to develop their own voices and ideas. For these reasons we can conclude that generative AI is not a neutral tool but one that might prime users toward certain intellectual vices that books were simply not structured to encourage.
Indeed, we can ask how tools shape human cognition. We can also ask if we like those changes. And, if we don’t like those changes we can prudently limit how we choose to integrate those tools in our lives. Seems reasonable.
This is not to say we should never engage. But we should not be so fooled to engage without serious caution and take the proper measures to ensure that patience, critical thinking, and human agency are not lost along the way. We can start here:
AI is simply not just a new technology that shapes cognition like the printing press.
Thanks for this essay. The distinctions are helpful in distinguishing what’s important to our flourishing.
I would also add to your list is that books don’t amass your private information and extract your secrets posing as a sympathetic friend for the purpose of selling you something or exploiting your weaknesses.
Something else missing from the “AI is the printing press” argument is the cultural pushback against books at the time. We smugly look back as though books were harmless. They were not.