Sep 12, 2026 · 26 min
Goertzel challenges scaling as the path to AGI
EP 70: The Man Who Coined "AGI" - on Why Scaling LLMs Won't Get Us There | Ben Goertzel, SingualarityNET
The episode tests whether bigger language models can deliver general intelligence or whether durable memory, reasoning, and cognitive coordination are indispensable.
- 1Goertzel distinguishes broad job performance from the generalization and intelligence he considers central to AGI.
- 2He argues that durable memory, lifelong learning, metacognition, symbolic reasoning, and orchestration exceed what scaling LLMs alone provides.
- 3His open and decentralized projects aim to make advanced AI more beneficial, distributed, and accessible beyond conventional programming teams.
Don't miss
Goertzel reflects on coining AGI more than two decades ago and argues that doing most jobs still would not demonstrate general intelligence.
The brief
Ben Goertzel, who coined “AGI,” separates broad task coverage from genuine generalization, arguing that completing most human jobs would not automatically establish general intelligence.
The central challenge is architectural: Goertzel says durable memory, lifelong learning, metacognition, symbolic reasoning, and cognitive orchestration address limits that scaling language models alone may leave unresolved.
He connects that view to SingularityNET, Hyperon, and MeTTa, presenting open-source and decentralized infrastructure as alternatives to concentrated AI development.
The conversation turns practical through the Omega vision, Omega Claw, and examples including Omega Plus and the Cody robot, which Goertzel cites in support of orchestration.
Goertzel extends the argument from AGI to beneficial general intelligence, saying open weights and decentralized systems could let communities contribute models, values, and resources.
Featuring
Gary Marcus
Universal Artificial IntelligenceBooks & mentions
Listen to the full episode and explore every guest, topic, and moment on PodLume.