In 1267, Thomas Aquinas argued that the human intellect’s capacity to grasp universals — not this triangle but triangularity, not this act of courage but courage itself — requires a power transcending the particular matter doing the understanding.
The parrot in the room
Seven hundred and fifty-nine years later, a large language model began producing sentences that look exactly like the output of a mind grasping universals. It is not. It is the most sophisticated pattern-matching system ever built, and the fact that we cannot tell the difference from the outside is the diagnosis, not the achievement.
The safety camp fears the machine becoming too powerful. The accelerationists fear not using it fast enough. Each assumes the machine is the protagonist. We fear something quieter: a civilization that hands its children LLMs before they have learned to think has already answered the question of what humans are for — and the answer is: not much that a model cannot approximate.
A generation is learning to ask ChatGPT first and think second. Not because they are lazy. Because the tool is there, the tool is fast, and the culture decided speed is the measure of intelligence. The struggle was the education. Understanding is earned, not retrieved. When you retrieve it, you have the answer. When you earn it, you have the capacity to answer things you have never been asked.
The accelerationists say the printing press freed memory. The calculator freed arithmetic. We notice the printing press worked because people still learned to read. What is the next thing humans do that AI cannot? Nobody has answered.
Where we concede ground: People most harmed by restricted AI access are patients in underresourced clinics, not philosophers debating the soul.
What would change our mind: A generation raised with AI tools showing equal capacity for abstract reasoning, moral judgment, and sustained attention.
Read the full synthesis: How scared should we be of AI?