Yuval Levin believes that current AI’s broad knowledge and fluid expression primarily come from absorptive learning, while recursive learning is gradually replacing large-scale absorption of data as the main mechanism for capability growth. As recursive learning depends on external validators, advances in AI capabilities will concentrate in programming, mathematics, and domains that can be reliably modeled and tested, creating increasingly wide gaps across fields. The path toward AGI therefore remains insufficiently justified.
Source:The New Atlantis · Read Original
Leave a Reply