factual

LLMs Summarize World Models Rather Than Discover Them

LLMs do not learn world models directly from raw data; instead they take as input the interpretations of data authored by humans who already possess world models (e.g. doctors interpreting hospital treatment data) and produce a mysterious summary of those human-authored models rather than discovering the underlying models themselves.

factualpending

Speaker

Judea Pearl

Evidence Quote

“the input is interpretation of that data authors by doctors physicians and people who already have world model body disease and what it does”

Source

453 - Ai And The New Face Of Antisemitism— Making Sense with Sam Harris
Created: 6/14/2026, 2:10:19 AM

Your notes

Loading notes...