Researcher: Real scientific progress isn't rewarded, expect decades of incrementalism

gerardsans · x · 2026-11-19

gerardsans argues the current AI research environment doesn't reward genuine scientific progress, so incrementalism may dominate for decades. He also questions why FFNs still use 2-4x feature expansion without full justification, and shares his attempts linking dead neurons to transformer behavior and informational capacity.

Original post →

More from AGI Musings

AGI Musings channel →