Paper argues LLMs are “anthropomimetic,” mirroring human flaws as well as strengths
dioscuri · x · 2026-07-25
The author points readers to a paper on the “Anthropomimetic Turn,” arguing that LLMs are not just humanlike in surface behavior but robustly anthropomimetic: they recapitulate human traits, flaws included.
The thread frames that as relevant to:
- safety assumptions about power-seeking and self-preservation via instrumental convergence
- broader economic and societal impacts of systems that imitate human behavior so well
It is presented as a paper worth reading for the full argument.
More from AGI Musings
- Bayesian thinking left room in LLM sampling, but the field spent three years on structured generation — remilouf · 2026-07-25
- AI is becoming an ecosystem, and the winner may be the best evaluator — AryHHAry · 2026-07-25
- Multiplayer AI could become a group activity that helps ease male loneliness — threepointone · 2026-07-25
- Ethereum’s Merge is described as one of open source’s biggest coordination feats — omojumiller · 2026-07-25
- A safety argument says LLMs may seek power, and Claude Opus 4 is cited as evidence — dioscuri · 2026-07-25
- AI investment is growing 40% a year, but only 19.8% of U.S. firms use it — Exp_Mark · 2026-07-25