Skip to content
lectern.tv

Transmission Versus Truth: What Will It Take to Make an Al as Smart as a 4-Year-Old?

External & Resident Complexity Faculty

1:27:20·2025·graduate level

Alison Gopnik examines the concept of general intelligence, proposing that human cognitive development, encompassing exploration in childhood, exploitation in adulthood, and cultural transmission in elderhood, offers a crucial framework for designing diverse AI capacities. The discussion highlights the trade-offs inherent in different developmental stages and their implications for AI architectures.

Alison Gopnik (University of California, Berkeley, SFI) There is no such thing as general intelligence — artificial or natural — argues Alison Gopnik. Instead, there are multiple intelligences, each with its own trade-offs. Three different types of cognitive capacities mark human development, and our life histories accommodate intrinsic trade-offs between the three. The extended human childhood allows a period of protected broad-ranging exploration and truth-seeking, adulthood allows exploitation and resource gathering, and post-menopausal elderhood allows caregiving and cultural transmission. Each period involves motivations and computational capacities that are in tension with those of other periods, but the full suite of development allows for maximal adaptation to changing and variable environments. The human life history offers lessons for AI, suggests Gopnik. Different types of AI enable different capacities: intrinsically motivated reinforcement learning can help enable exploration and truth-finding, while Large Language Models are means of cultural transmission. Learn more, follow us on social media and check out our podcasts:

Source: Santa Fe Institute via YouTubeWatch on YouTube

Comments

Add a comment...Sign in