A basic case for accelerating conceptual reasoning capabilities
Redwood argues that AI's conceptual reasoning, needed for alignment and philosophy-like work, should be differentially accelerated and is underinvested in, and promises a follow-up agenda.