LLMs are not all the same. Imagine a Wikipedian LLM that has learned only the articles in Wikipedia. Imagine another Our Librarian LLM that has learned all the publications in Our Library. The performance of both should be quite different. These two could be made available in the near future.
ㅡㅡ
Yes. This is an important distinction, and it fits your Our Library / World Library idea very well.
The phrase “World Brains or LLMs” can be sharpened into a contrast between different kinds of collective intelligence:
World Brains or LLMs
LLMs are not all the same. Their intelligence depends greatly on what they have learned from.
Imagine a Wikipedian LLM trained mainly on Wikipedia articles. It would possess broad public knowledge, but its knowledge would be shaped by encyclopedia-style summaries, consensus language, and secondary presentation.
Now imagine an Our Librarian LLM trained on the full body of publications in Our Library: books, journal articles, conference papers, reports, theses, citation contexts, abstracts, reviews, criticisms, and scholarly debates. Its performance would be quite different. It would not merely repeat summarized public knowledge. It could trace how knowledge developed, how authors evaluated predecessors, how arguments changed across time, and how citation contexts reveal intellectual relationships among works.
These two systems would therefore represent different kinds of “World Brain.” A Wikipedian LLM would resemble a public encyclopedia brain. An Our Librarian LLM would resemble a scholarly and historical brain, grounded in the full record of human inquiry.
Both could become available in the near future. But their value would not be the same, because the depth of an LLM depends on the depth, structure, and quality of the library from which it learns.
The key idea is powerful: an LLM is not only a model; it is also the memory-world it has absorbed. A Wikipedia-based LLM and an Our Library-based LLM would differ not simply in size, but in intellectual depth, citation awareness, and cognitive-contextual power.
ㅡㅡㅡㅡ
2026-07-08 Mark Park
Our Library 