About one-fifth of daily life is already mediated by AI, from social media feeds to dating apps, and this share is growing rapidly as AI becomes more capable. The speaker uses Socrates’ ancient critique of writing and books to frame a central concern: that AI, like books before it, risks creating a dangerous dependency that erodes our own faculties for judgment and truth-seeking.
Socrates’ critique of writing
Socrates, as reported in Plato’s dialogue Phaedrus, argued that writing and books appear to enhance wisdom and memory by serving as a “second brain” we can always reference, but in reality they weaken our own capacity to internalize knowledge and think for ourselves.
Books cannot respond dynamically to the reader or meet them where they are; they say the same thing to everyone.
By offloading knowledge to an external medium, we try less to deposit it in our own minds.
The speaker believes Socrates was fundamentally right: something genuine has been lost with the invention of books.
What was lost: oral culture and forced memorization
Oral cultures required extraordinary feats of memory, such as the centuries-long oral transmission of Homer’s Iliad and Odyssey before they were ever written down.
Modern people find it nearly impossible to memorize and recite works of that scale, suggesting a real cognitive loss.
A similar pattern appears in education: Shakespeare’s original audiences could follow his complex language because forced memorization of the classics was a standard part of education, building a poetic sensibility that cannot be achieved without that effort.
This mirrors the concern in the workplace today: junior lawyers and bankers who no longer perform tedious foundational tasks because AI handles them may never develop the deeper professional judgment those tasks cultivated.
What looks like a task easily automated away turns out to be the basis for a more important faculty.
The dual nature of technology
Books illustrate a general pattern with powerful technologies: they elevate us in some ways while creating dependency that degrades our own capacities in others.
Books enabled vastly faster accumulation and exchange of knowledge than oral culture alone could achieve, especially after the printing press.
Socrates’ critique does not mean we should abandon books, but it highlights the tradeoff: power gained versus faculty lost.
The same dual aspect now applies to AI. It can save us tremendous tedium and help us arrive at better decisions, but if it becomes an “autocomplete for life,” it risks castrating our own ability to seek truth and exercise judgment.
Plato’s dialogue form as a partial solution
Plato preserved Socrates’ critique of books in a book of his own, which raises an apparent contradiction.
One interpretation is that Plato agreed with Socrates and therefore did not value his written dialogues highly, preferring one-on-one tutoring, as with Aristotle.
A more interesting reading is that Plato deliberately chose the dialogue form as a superior way to write books that addresses Socrates’ concerns.
Platonic dialogues often end in aporia—Socrates himself is left without a definitive answer to questions like “What is justice?” or “What is virtue.”
Rather than handing down fixed propositions on a stone tablet, the dialogue form invites readers to wrestle with the questions themselves, preserving the active exercise of judgment.
Plato thus harnessed the power of the written medium while attempting to combat the dependency it creates.
The challenge for AI
The central question for AI design is: what form of AI, what product, interface, or user behavior can harness AI’s power without making us completely dependent on it and eroding our truth-seeking ability?
One idea is a Socratic chatbot that prods the user with questions rather than simply providing answers.
Another is a marketplace of chatbots in open debate, letting users see multiple perspectives and form their own judgments.
The speaker does not claim to have the answer. Cosmos, in partnership with Fire, is offering $1 million in no-strings-attached grants for prototypes of “truth-seeking AI” that address this challenge.
