Beyond "stochastic parrots": Wigner's unreasonable effectiveness now applies to LLMs
anselm · x · 2026-09-17
Sharing Scott Barry Kaufman's argument against dismissing LLMs as "stochastic parrots".
Kaufman invokes Eugene Wigner's 1960 essay on the unreasonable effectiveness of mathematics — math invented purely for elegance kept describing the physical world. He contends the same unreasonable effectiveness has now reached human domains: language, conversation, reasoning, and the mind. Nobody designed next-token prediction as a theory of thought, yet LLMs comprehend real human interactions and reason as well as or better than most humans. The thread anticipates the "this isn't surprising" objection and opens a deeper debate on whether LLMs truly understand.
More from AGI Musings
- Full video: Kunal Shah on CRED's 90% AI-written code — rohanpaul_ai · 2026-09-17
- CRED founder Kunal Shah: 90% of our code is now written by AI — rohanpaul_ai · 2026-09-17
- Marc Andreessen: AI Safety panic is just the old guard clinging to soft power — beffjezos · 2026-09-17
- Most people are vessels for ideas, like neurons in a larger mind — nptacek · 2026-09-17
- Musk Often Right on Outcome, Wildly Off on Timelines, Dev Says — geoffwolfe · 2026-09-17
- Oxford's Sandra Wachter on likelihood of an AI slowdown, cooperation and accountability — SandraWachter5 · 2026-09-17