[X]score: 0.25
Gary Marcus: Scaling Detour Cost AGI Progress, Neurosymbolic Path Remains
May 27, 2026
Gary Marcus argues that over-investment in pure scaling delayed three of four AGI prerequisites he identified in 2020 — structured knowledge databases, reliable reasoning under uncertainty, and explicit world models — while neurosymbolic AI (e.g., Claude Code, DeepMind's math system) has partially vindicated his first goal.
HOW THIS AFFECTS YOU
●
researcherWorth watching as a framing argument: if neurosymbolic + world models + structured knowledge is the actual path, it signals where to direct research effort beyond scaling.
●
founderIf the scaling plateau narrative gains traction, product bets on reasoning, knowledge graphs, and world models may be better positioned than pure LLM wrappers.