Two hundred years ago, only 12 percent of adults worldwide could read. Today that figure sits at 87 percent—a shift driven not just by education policy but by actual biology. As billions acquired literacy, their brains physically rewired: neural connections between hemispheres thickened, regions evolved for face recognition got repurposed for letter recognition, and entirely new pathways activated in response to written language. No genes mutated. Pure cultural practice reached inside the skull and reorganized the organ that makes us human. Now apply that logic to AI—and start worrying.
The Next Great Rewiring Has Already Started
A new analysis from Baillie Gifford investment manager Fraser Thornton argues that AI represents the most pervasive cultural technology since literacy—spreading faster than anything in human history and already reshaping cognition at scale. 'This is not an exception but the rule,' Thornton writes, drawing on Joseph Henrich's research into how cultural technologies from cooking to markets have systematically reshaped human physiology and psychology. The Catholic Church spent a thousand years controlling variation (narrowing orthodox thought), dominating transmission (near-monopoly on literacy and education), and rewiring selection (dismantling kinship structures through enforced monogamy). AI is doing all three simultaneously, at unprecedented speed.
Killing the Apprenticeship Model
The most immediate damage is structural. Junior lawyers, accountants, and doctors have traditionally built competence through grunt work supervised by senior practitioners—tedious but real learning that develops judgment through failure. When Shopify's CEO recently told teams they must first prove AI cannot do a task before requesting headcount, he crystallized what Thornton calls 'the efficiency gain and the training loss are the same decision viewed from different angles.' Entry-level hiring at major tech companies has already fallen more than 50 percent below pre-pandemic levels, hollowing out career pathways that once trained entire generations of professionals.
The Metacognition Trap
Research published in Computers in Human Behavior found that AI use improves performance while degrading metacognitive accuracy—users become better at producing outputs but worse at evaluating whether those outputs are correct. Thornton observes this directly: 'I have watched my own cognitive habits change in ways I did not choose and barely noticed.' The investment manager holds stakes in the companies building these tools, believes in their potential, and still watches his own thinking shift without conscious consent. If it's happening to someone whose entire career involves analyzing these forces, it's happening to everyone.
Cultural Transmission at Scale
When a child asks ChatGPT why the sky is blue, they are learning from a single model trained on accumulated human text—not any individual teacher. That model becomes cultural instructor to hundreds of millions simultaneously, transmitting a more centralized body of knowledge than any institution in history. Research in Nature Human Behaviour found that AI prompted in Chinese versus English exhibits systematically different cultural orientations: interdependent and holistic in Chinese, independent and analytic in English. Since most training data comes from English-language sources rooted in individualistic western cultures, users in collectivist societies absorbing AI-mediated answers may adopt psychological norms without realizing it.
The Selection Environment Problem
Recommendation algorithms have become the dominant selection mechanism for cultural content—determining which news reaches millions, which artists find audiences, which political arguments gain traction. These systems don't select for truth or usefulness. They select for engagement metrics that drive advertising revenue. Thornton argues this creates 'a selection environment that favors certain kinds of cultural traits over others, not because those traits are adaptive in any meaningful sense, but because they happen to align with the metrics.' The diversity of human thought is narrowing in real-time.
What Actually Matters
Pilots still learn manual flight before autopilot—not because automation fails often, but because someone needs to land the plane when it does. An accountant who prepared hundreds of returns by hand spots errors buried in AI-generated filings. A doctor who made diagnoses without AI support can override confident wrong predictions. Thornton's conclusion: 'The people who will thrive are not those who use AI the most, but those who can still think without it.' Professional bodies and universities need to preserve training pathways that build genuine expertise even when AI makes them look slow—then teach AI orchestration as an advanced capability sitting on top of that foundation.
The Bottom Line
The Catholic Church took centuries to rewire European psychology. Literacy took generations to reshape the brain. AI is doing both at once, to billions, in years—and we're not moving fast enough to preserve what makes us cognitively irreplaceable. Watch your own thinking. It's already changing.