Despite their immense processing power, todays most sophisticated artificial intelligence systems still fall far short of toddlers when it comes to learning language, according to new research.
In fact, researchers say that if a human learned language at the same rate as ChatGPT, it would take them 92,000 years. This stark divide is the focus of a newly published framework in Trends in Cognitive Sciences by Professor Caroline Rowland of the Max Planck Institute for Psycholinguistics and colleagues at the U.K.s ESRC LuCiD Center.
Synthesizing findings across neuroscience, linguistics, psychology, and computer science, the research explores why young children so consistently outperform AI at acquiring natural language — and what that tells us about both human development and the limits of machine learning.
The researchers argue that the difference isnt just about how much data each learner receives. Its about how that data is processed. AI systems, including large language models like ChatGPT, learn passively from massive quantities of written text. Children, by contrast, engage with the world in a dynamic, embodied way. They learn language through direct experience — by hearing it, speaking it, and linking it to everything they see, touch, and do.
“AI systems process data … but children really live it,” said Rowland. “Their learning is embodied, interactive, and deeply embedded in social and sensory contexts. They seek out experiences and dynamically adapt their learning in response — exploring objects with their hands and mouths, crawling towards new and exciting toys, or pointing at objects they find interesting. Thats what enables them to master language so quickly.”
Recent advances in research tools, from AI-enhanced speech recognition to wearable eye trackers, have made it possible to study childrens learning environments in unprecedented detail. But Rowland and her team say theoretical models of language acquisition have lagged behind the data. Their new framework aims to close that gap by focusing on how children actively construct meaning from richly layered, real-world experiences.
These findings dont just reshape how we understand child development — they also pose provocative questions for the future of AI. If we want machines to learn language the way humans do, Rowland suggests, we may need to build them to interact with the world more like children.
“AI researchers could learn a lot from babies,” she said. “If we want machines to learn language as well as humans, perhaps we need to rethink how we design them — from the ground up.”