Kids outlearn AI—and we still don’t know why

The article examines the data efficiency gap between human children and large language models, with children learning language fluency from approximately 100 million words while modern LLMs require trillions of tokens. Researchers are conducting experiments through initiatives like BabyLM and training models on egocentric video footage to understand how machines might learn language more efficiently and what this reveals about human language acquisition.

Source: MIT Technology Review