Reading up on continual-learning

Dwarkesh Patel argues that current AI models lack the continual learning and on-the-job adaptability of humans, making AGI distant, but he expects explosive progress once true AGI arrives within a decade or two.