causal

Why LLMs don't make novel discoveries from combined knowledge

LLMs haven't made novel scientific discoveries by combining their vast knowledge not because of a fundamental limitation but because no one has seriously tried to build the scaffolding, scale the model, or — most importantly — train them to do it, since the pre-training process doesn't incentivize this kind of connection-making.

causalpending

Speaker

Daniel Kokotajlo

Evidence Quote

“did the training environment train it to do this? And often the answer is no. And often I think that's a good explanation for why the AI is not good at it”

Source

AI 2027: month-by-month model of intelligence explosion — Scott Alexander & Daniel Kokotajlo— Dwarkesh Patel Podcast
Created: 6/13/2026, 3:33:52 AM

Your notes

Loading notes...