LLMs feel von Neumann-coded, not Einstein-coded: reflections on the AI narrative problem
akbirthko · x · 2026-08-25
Citing a prescient graph from the classic blog Wait But Why in light of recent progress in theorem proving, the author argues LLMs still feel very "Roman" and von Neumann-coded rather than "Greek" and Einstein-coded — dense computation rather than revolutionary insight.
In the thread, they add that there's a massive optics problem: people talk about LLMs the way Freeman Dyson described the "technical arrogance" around fusion bombs, with a weird Oppenheimer fetish. They call for a more life-affirming way of discussing these systems — "they're not nukes" — closing with Moravec's advice to "build arks and adopt a seafaring life."
Related event: Debate: Should LLMs Be Compared to Nukes or von Neumann?(2 posts)→
More from AGI Musings
- AI Drives First US Grid and Blue-Collar Job Expansion in 20 Years — danielrock · 2026-08-25
- Open source models rapidly closing the gap with closed source AI — iScienceLuvr · 2026-08-25
- Data center fights prelude future battles on reshoring input-heavy, job-light manufacturing — bradneuberg · 2026-08-25
- Stanford study: AI hits entry-level jobs hardest — GarlicoinAccount · 2026-08-25
- AI energy spend will be industrial, leaving abundant compute for individuals — tawnniee · 2026-08-25
- The Inevitability of the Company AI Harness — MxMnr · 2026-08-25