Pragmatics in language acquisition and AI systems

Cognitive scientist Michael Frank studies differences in how children and AI learn language. There is a “data gap” between the billions of words ChatGPT has to work with and the millions of words a toddler is exposed to. But, says Frank, children learn in a rich social context that supports their learning. He’s currently conducting the “BabyView Study,” where he puts cameras on young children's heads to help him understand their learning experience, as Frank tells host Russ Altman on this episode of Stanford Engineering’s The Future of Everything podcast.

Episode Reference Links:

• Wordbank: An open database of children's vocabulary development
• ManyBabies
• The SAYCam
• BabyView Study

Connect With Us:

• Episode Transcripts >>> The Future of Everything Website
• Connect with Russ >>> Threads / Bluesky / Mastodon
• Connect with School of Engineering >>> Twitter/X / Instagram / LinkedIn / Facebook

Chapters:

(00:00) Introduction

Russ Altman, host, introduces the episode's focus on the future of language learning and guest, Michael Frank, an expert on how children acquire language

(02:05) Child Vs AI Language Acquisition

Child language acquisition and the comparison with AI systems like ChatGPT, the differences in language learning between humans and AI, including the vast data exposure AI systems have compared to children.

(05:23) Data-driven Approaches

The importance of big data in understanding language acquisition and the Wordbank database.

(08:04) The Early Stages of Language Learning

The universal aspects of language learning across different languages and cultures, specifically babbling and its evolution into language-specific sounds.

(10:30) Exploring Multimodal Language Learning

Introduction to the SAYcam project, and the BabyView study both using camera footage from children to gather multimodal information. 

(13:12) Social Learning and Language Acquisition

Insights into how social interactions and grounded contexts facilitate children's language learning. Exploration of active learning strategies in children.

(18:22) The ManyBabies Project

A Global research consortium to understand child development across cultures, their first study documented infants' global preference for infant-directed speech. Michael outlines the funding challenges

(21:28) Understanding Pragmatics and AI Common Sense

Explanation of pragmatics in language acquisition and its importance in social cognition and inference making. The connection between pragmatic language skills in children and the challenge of instilling common sense in AI systems.

(24:21) The Role of Reading in Language Learning

The benefits of reading to children for language development while still recognizing diverse learning paths and early auditory learning.

(27:54) Multilingualism and Early Learning Flexibility

The natural ease of learning multiple languages in childhood  and when and why the flexibility in language learning begins to decline.

(30:35) Conclusion

Connect With Us:

Episode Transcripts >>> The Future of Everything Website

Connect with Russ >>> Threads / Bluesky / Mastodon

Connect with School of Engineering >>>Twitter/X / Instagram / LinkedIn / Facebook