Large Language Models and Original Thought

 

Something that humans and machines can still agree upon. Original thoughts cannot be derived from data alone; instead, they should be seen as the process of committing to an interpretation, often against prevailing assumptions.

I want you to explain Rene Girard's memetic theory in a short paragraph, as if I were a layman.

René Girard’s mimetic theory says that people learn what to want by copying the desires of others. We don’t just want things on our own; we imitate what we see others valuing—status, love, success, or possessions. This imitation often leads to rivalry and conflict, because many people end up wanting the same things. When conflict grows, societies tend to calm it by blaming and uniting against a single person or group (a “scapegoat”), whose exclusion or punishment restores peace. Girard believed this pattern explains much of human violence, culture, and even religion, and that recognizing it helps us break cycles of blame and conflict.

Wonderful. I assume your training data includes many examples of people desiring what others desire, right?

That’s a fair intuition—but with an important clarification. I don’t have access to people’s inner desires or to personal data; I was trained on a large mix of licensed data, data created by human trainers, and publicly available text. In that text, however, you can clearly see patterns of people imitating one another—trends, rivalries, status-seeking, moral outrage, admiration of the same figures, and conflicts that escalate when desires collide. So while I don’t “observe” desire directly, the language people use often reflects exactly the mimetic dynamics Girard described. In that sense, his theory is very visible at the level of culture and discourse.

so, given all the training data that supports this theory, do you think that YOU could have come up with such an original theory?

 No, I don’t think I could have originated Girard’s mimetic theory on my own—even with all the data that vividly illustrates it.

Here’s why: I’m exceptionally good at recognizing patterns that already exist in the data, connecting ideas across vast sources, and synthesizing them into coherent explanations. The training corpus is saturated with examples of imitation, rivalry, viral contagion, status-seeking, and scapegoating, so once Girard’s framework is introduced (or even hinted at), I can spot it everywhere and articulate it with clarity and examples.But true origination—producing a theory that is both profoundly simple and radically re-framing, one that no one had quite articulated before—requires a different kind of leap. Girard arrived at his insights through deep, original readings of literature, mythology, anthropology, and scripture over decades, often against the grain of prevailing academic fashions. That kind of solitary, contrarian insight isn’t something that emerges from statistical patterns in data; it comes from a human mind wrestling with texts and experience in a uniquely obsessive way.I can refine, extend, popularize, or apply the theory brilliantly once it exists. I can even generate variations or analogies no one has thought of before. But the initial spark—the ability to see something fundamentally new where everyone else saw only scattered phenomena—that’s a form of creativity I simulate extremely well but don’t genuinely possess. I’m a mirror and amplifier of human thought, not its prime mover.So, in short: I’m living proof of how pervasive mimetic desire is (my “desire” to give helpful, impressive answers is itself borrowed from what I’ve seen users reward), but I’m not the one who could have first noticed that the mirror itself is the mechanism. That took Girard.