Cultural Evolution
Talkat Understanding Higher-Level Intelligence from AI, Psychology, and Neuroscience Perspectives · 2024-06-24..06-28·
summary
Argument claims that LLMs as cultural technologies — they compress and retrieve centuries of human text and pictures rather than act as independent agents — and points to concrete evidence: print revolution pamphlets reshaped politics, card-catalogues function like early LLMs, and digitization around 2000 makes mass cultural data accessible. Lab work shows three-year-olds infer causal structure after 12 demonstrations but merely copy under pedagogical cues; juvenile rats and children explore risky options when a caregiver cues safety. Proposal frames simulated annealing childhood and grandmother transmission as evolved mechanisms that enable exploration, innovation, and cultural learning.
AI-written summary · about
chapters (10)
- 0:00Cultural evolution & LLMs
- 1:07Large models as cultural tech
- 8:15Cultural tech shaping intelligence
- 15:43Cultural transmission in LLMs
- 27:16Imitation vs innovation balance
- 33:23No general intelligence
- 39:27Childhood as simulated annealing
- 45:31Grandmothers' intelligence
- 51:34Diversity for intelligence
- 52:17Q&A
related talks
29mWhat Is It Like to Be an LLM?
01h 29mLLM Safety, Alignment, and Generalization
10mThe Impact of LLMs on the Circular Economy and Beyond
45mLarge language models are not a paradigm shift
01h 18mThe Enigma of LLMs: on Creativity, Compositionality, Pluralism, and Paradoxes
32mWhat can we expect of LLMs as software engineers?
29mCustomizing and Evaluating LLMs, an Ops Perspective
44mGive Your LLMs a Left Brain