MIT Technology Review revisits a long-standing observation about language learning: across an estimated 100,000 years of human communication, only human children could acquire a language to complete fluency. The article notes that this is no longer the only example, pointing to the arrival of AI language systems in the years since ChatGPT’s release roughly four years earlier.
The framing highlights a gap in understanding: even as AI has become capable of handling human language, children still outperform these systems at learning it, and the reasons for that advantage remain unexplained.
Why it matters
Comparing how children and AI systems acquire language touches on foundational questions about learning itself. The piece signals that the mechanisms behind children’s superior language learning are not yet understood, which is relevant to how AI language models are studied and developed.