Ideas / Dwarkesh: Arguments About the AI Trajectory
Is RL + LLMs Enough for AGI?
Dwarkesh Patel with Sholto Douglas and Trenton Bricken
Sholto Douglas and Trenton Bricken look beneath the forecasts at reinforcement learning, long-horizon agents, continual learning, and interpretability.
Executive predictions are cheap; reinforcement-learning environments are not. Douglas and Bricken discuss what it would take to turn language models into systems that can pursue long tasks, learn continually, and contribute to autonomous software-engineering work.
Their mechanistic perspective is a useful test for every confident date elsewhere in this collection. The calendar still has to compile.