causal

Autoregression Is Ill-Suited to Unlikely Connections

Autoregressive next-token generation is a strange way to produce insight, because the valuable connections are by nature unlikely next tokens, and standard RL does not specifically upweight such low-probability leaps, so the system may hold intelligence locked inside an awkward interface.

causalpending

Speaker

Grant Sanderson

Evidence Quote

“What’s the thing that’s specifically upweighting and incentivizing making these unlikely connections when the vast majority of them aren’t the predictable next token?”

Source

Dwarkesh Patel Podcast - Grant Sanderson – AI and the future of math— Dwarkesh Patel Podcast
Created: 11/17/2026, 6:11:54 AM

Your notes

Loading notes...