Have you ever wondered if the whirring circuits of an AI like ChatGPT process language in a way similar to our brains? A fascinating new study suggests a surprising commonality: both humans and large language models (LLMs) seem to rely on hidden hierarchical structures, like constructing a sentence with building blocks of meaning. Researchers devised a clever "word deletion" game where participants had to infer a pattern from a single example and apply it to a new sentence. The twist? The deleted parts were always grammatical building blocks known as "constituents." The results? Both humans and LLMs like ChatGPT showed a strong preference for deleting these constituents, hinting at a shared underlying mechanism for understanding sentence structure. This wasn't just about memorizing patterns. A simpler AI model that only looked at word order and individual word properties couldn't crack the code as effectively, especially with limited examples. Interestingly, the study also revealed that the specific "rules" inferred for the game varied between Chinese and English, suggesting that the way we process language might be influenced by the unique structures of different languages. This research opens a window into the hidden workings of both human and artificial minds, suggesting that even with vastly different architectures, we might share some fundamental strategies for making sense of language. While LLMs are trained on mountains of text, humans learn language through a richer, multi-sensory experience. Yet, this study suggests that both arrive at a similar outcome: a latent understanding of how words fit together to create meaning. This discovery has exciting implications for building more transparent and human-like AI. By understanding the shared principles underlying language processing, we can potentially bridge the gap between artificial and human intelligence, paving the way for more intuitive and effective communication between humans and machines.