Why kids learn language faster than artificial intelligence

Your five-year-old learns to speak by listening to conversation around the dinner table. An artificial intelligence system? It needs to process hundreds of thousands of times more examples to learn the same language skills.

This gap is the subject of growing scientific interest. Large language models (LLMs)—the AI systems that power tools like ChatGPT—work by analyzing enormous amounts of text data to predict what word should come next in a sentence. They're incredibly effective, but the amount of data they require is staggering compared to how children learn.

What makes this puzzling is that we still don't fully understand why kids are so much better at this. Children seem to learn language through genuine understanding and context—picking up meaning from the world around them, not just patterns in text. AI systems, by contrast, work through pure pattern recognition. They find mathematical relationships in data without actually understanding what words mean.

For everyday users, this research matters because it hints at a real limitation in how today's AI learns. If we could figure out why human brains are so efficient at learning language, we might be able to build smarter, more capable AI systems that don't need to consume the entire internet to understand a conversation.