LLMs Like ChatGPT Match Human Fluency but Demand Vastly More Data

Large language models (LLMs) such as ChatGPT have reached near-fluent human language proficiency, a milestone previously unique to humans, according to MIT Technology Review. While human children typically master language within about four years, LLMs require processing roughly a hundred thousand times more words to achieve comparable fluency.

MIT Technology Review highlights that humans have been communicating for at least 100,000 years, with children naturally acquiring language through interaction and exposure. In contrast, LLMs like ChatGPT, Claude, and DeepSeek rely on enormous datasets to learn, reflecting an inhuman scale of data consumption compared to human learning.

For Japanese market participants, understanding the data demands behind LLMs is crucial as AI-driven language tools increasingly influence FX, crypto, and equity trading strategies, where rapid and accurate language processing can provide a competitive edge.