AGI debate: pretraining won because the economy is PDFs, but LLMs still can't learn
akbirthko · x · 2026-11-12
Victor Taelin lays out his AGI thesis: any reasonable definition of AGI implies a system that learns as efficiently as a human, and LLMs clearly fail that bar — weights don't change during conversation, so they learn nothing. Without true infinite context or 10,000x cheaper GPUs, current methods won't scale to AGI in that sense, even if future models become capable of building Dyson spheres. Reposter akbirthko adds his take: pretraining succeeded because the global economy is PDFs, a fact neither Sutton nor LeCun respected.
More from AGI Musings
- Clara Shih: junior workers face the worst job market in 37 years — clarashih · 2026-11-12
- RLHF Co-creator: LLMs Are Built for Assistance, Picking the Right Task Is Key to Automation — FinanceYF5 · 2026-11-12
- He gave up Harvard Law and an Anthropic offer to fight AI takeover — and got mocked — AaronBergman18 · 2026-11-12
- Vertex exec calls AI-cures-all claims "complete BS": 50 AI firms, zero working drugs — AlexTensor · 2026-11-12
- tinygrad slams EA: 'any means justify their ends' culture doomed to an SBF-style blowup — mark_k · 2026-11-12
- A painter asks: when an AI image moves you, where does the meaning live? — indianbull88 · 2026-11-12