The economy doesn't need true AGI, says Replit CEO (www.businessinsider.com)

Replit CEO Amjad Masad told the a16z podcast that the industry doesn’t need a human‑like, god-tier AGI to transform the economy — “functional AGI” will suffice. By his definition, functional AGI are systems that learn from real‑world data and can complete verifiable tasks autonomously without requiring consciousness or human‑style reasoning. Masad argued these systems are already within reach and could automate large swaths of labor across sectors, making them economically and socially transformative even if a true general intelligence remains elusive. Masad also warned the field may be stuck on a “local maximum,” optimizing incremental gains on large models rather than pursuing radical new architectures needed for true AGI. His view reflects a growing debate: major labs still prize AGI, but critics like Gary Marcus and even some researchers (Yann LeCun) argue scaling alone won’t get us there and breakthroughs may be decades away. The practical implication for ML practitioners and businesses is clear — focus on deploying robust, data‑driven systems that deliver verifiable automation today, while acknowledging that the lofty goal of general intelligence may require fundamentally different approaches.