causal

LLMs Work by Narrowing the Search Space

LLM-based AI and chess engines are improved by reducing options—chopping off undesirable parts of the search space via heuristics and tweaking out hallucinations—whereas an AGI requires the opposite: ensuring no possible avenue of thought is closed off to it.

causalpending

Speaker

David Deutsch

Evidence Quote

“for an agi the difficulty is how to make it so there is no possible avenue of thought that is close to it”

Source

414 - Strange Truths— Making Sense with Sam Harris
Created: 6/14/2026, 2:03:47 AM

Your notes

Loading notes...